1
The Philosophy of
Diseases and Causes
DRAFT MANUSCRIPT
Benjamin T. H. Smart
University of Johannesburg, South Africa
ben@jerseyserve.com
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
2
For MJT
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
3
Contents
Acknowledgements
List of abbreviations
Introduction
1. The Concept of Disease in Clinical Medicine
- The maximally value laden conception – Rachel Cooper on disease
- The pure statistical conception
- The frequency and negative consequences approach, and the line-drawing problem
- The etiological account of function, and disease as harmful dysfunction
- Disease as harmful function – ‘drawing the line’ on the etiological account of disease
2. What is a Pathological Condition?
- Boorse’s naturalism
- Objections to Boorse’s naturalism
- The frequency and negative consequences approach revisited
- The etiological theory of pathological condition
3. Concepts of Causation in the Philosophy of Disease
- Causation as counterfactual dependence
- Clinical medicine and the dispositional account of causation
- The classification of diseases, and the sufficient-cause model of causation
4. Causal Inference in Public Health
- Hill’s criteria and the evidence-based medicine evidence hierarchy
-The epidemiologist’s potential outcomes approach
- Hernan and Taubman’s potential outcomes approach
- Diffusing Broadbent – a Popperian take on the potential outcomes approach
- The importance of nonmanipulable causes
5. Concluding Remarks
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
4
Acknowledgements
I am grateful to Kate Graham, Alex Broadbent,Pendaran Roberts, and especially to Michael
Talibard.
Section 4 of chapter three is provided with kind permission from Springer Science+Business
Media. Smart, B.T.H.On the Classification of Diseases.Theoretical Medicine and
Bioethics(August 2014);35(4) pp.251-69
Figures 2.and 4. are provided with kind permission from The University of Chicago Press.
Schwartz, P. H. Defining Dysfunction: Natural Selection, Design, and Drawing a Line.
Philosophy of Science(July 2007): p.272 and p. 378 © The Philosophy of Science Association.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
5
List of Abbreviations
BMI
Body mass index
BST
The biostatistical theory
CCD
Causal classification of disease
EBM
Evidence based medicine
EBP
Evidence based practice
ETPC
The etiological theory of pathological condition
GP
General Practitioner
FNC
The frequency and negative consequences approach
NHST
Null hypothesis significance test
POA
Potential outcomes approach
RCT
Randomised control trial
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
6
Introduction
Disease is everywhere. Everyone experiences disease, everyone knows somebody who is, or
has been diseased, and disease-related stories hit the headlines on a regular basis. Just this
morning, in fact, I heard thatbeing tall increases one’s risk of cancer. I am not a short man,
so this did not please me. On the other hand, thirty seconds later, the same news reader
announced that the key to curing cancer might just be found in the sting of an unusual
species of wasp – so I left the house relatively content.
Unfortunately, when diseased, one often suffers pain and nausea, spends more time
than usual in the bathroom(!), and is generally unable to do the things one enjoys.
Eventually, of course, it is disease1 that relieves us of our mortal coil.
Fortunately, if one is diseased, one can generally do something about it; namely, one
can either go about treating oneself (perhaps by taking a pain killer), or when more serious,
one can visit a clinician. If one is lucky, the clinician will be able to explain what is causing
the symptoms, diagnose a condition, and prescribe treatment. Regularly, however, a single
visit to your general practitioner (GP) is not enough for a diagnosis, since she will often have
to send blood samples/urine samples/skin samples etc. for analysis. Here, the clinician asks
for the pathologist’s help.
Unlike the GP, the pathologist knows nothing of how the patient is feeling, or what
the patient’s future intentions are. Those qualities are relevant to the clinician, since she
must decide how to proceed with treatment, but not to the pathologist. Very crudely, the
1
In the ‘pathological condition’ sense.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
7
pathologist’s job is to identify the disease a patient is suffering from, by carefully
testing/examining the physical evidence at her disposal (and on occasion, to determine
what caused a patient’s death).
Once the clinician can make a diagnosis (based on the pathologist’s conclusions), she
can prescribe treatment - but she would be unable to do this, without the vast quantity of
medical research that occurs ‘behind the scenes’; the clinician must know what the
appropriate treatment is. Establishing the determinants of, and treatments for disease,
involves designing and conducting all kinds of studies and trials, and doing lots of statistical
analysis. This is the task of epidemiologist. Crudely stated, epidemiology is “the study of
health and disease in populations” (Saracci, 2010, 2), for the purposes of improving public
health. Without it, clinical practice (at least as we know it today) could not get off the
ground.
Unsurprisingly, much of the philosophy of disease literature targets a conceptual
analysis of disease. However, there is more to the philosophy of disease than answering the
question: ‘what is disease?’ Clinical medicine, pathology, and epidemiology, all carry with
them a host of interesting philosophical issues. The target of this book is to provide the
reader with an insight into some of the existing arguments in the philosophy of disease, and
to build on these debates where I can.
In part one I follow the trend, and consider what we mean when we talk of ‘disease’.
This question cannot, however, be answered from an ‘in general’ perspective, since as we
have just seen, medicine comprises several, very different disciplines. I therefore analyse
the concept of disease, first from the perspective of the clinician, and then from the
perspective of the pathologist. In this short book there is not space to consider every
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
8
argument in the literature. However, by focusing on the most prominent figures in this
debate, I hope to provide the reader with a good understanding of the more influential
views, and their problems. I also propose two novel conceptions of disease, one for the
clinician, and one for the pathologist. We shall see that the concept of the clinician and that
of the pathologist, do not coincide – the answer to the question ‘what is disease?’, is thus
entirely context dependent.
In part two of the book I discuss causation in medicine. Since Hume, numerous
analyses of causation have been offered by analytic philosophers. I apply several of these to
medicine, discovering that those working in the medical professions use different
conceptions of causation in different circumstances. Interestingly, we shall see that some
conceptual analyses of causation are useless for clinical practice, but nonetheless crucial for
improving public health. To end chapter three, I propose a model for classifying diseases by
their causes.
The final chapter of this book concerns causal inference in public health. In this
section, I analyse two theoretical notions employed by epidemiologists: the potential
outcomes approach to effect measurement, and Kenneth Rothman’s sufficient-cause model
to epidemiology. The philosophy of epidemiology is (at the time of writing) a fledgling
discipline. I hope this discussion will provoke more research into a hugely underexplored
area of philosophy.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
9
Part One.
The Concept of Disease
Chapter 1. The Concept of Disease in Clinical Medicine
1. Introduction
[H]ow we think of health and disease lies at the very core of medical practice,
reflections on bioethics, and the formation of health care policy’ (Engelhardt in
Ananth 2008).
This chapter asks the question: ‘what, for the clinician, is disease?’ Of all the tasks
philosophers undertake, I consider the conceptual analysis of disease to be of particular
importance – and for obvious reasons. To mention just a few: concepts of health, illness and
disease (specifically mental illness) affect whether or not one is legally responsible for one’s
actions; a concept of disease is necessary to demarcate the diseased from the healthy
(which might distinguish those eligible for medical treatment, from those not2); not only
must medical care be paid for, but one’s state of health affects one’s right to disability
allowance, free housing, and so on - so concepts of health and disease have a significant
economic impact; and since the decisions made by those in the medical profession
affectlegal responsibility, eligibility for medical care, eligibility for income support, and the
distribution of resources (both locally and internationally), decisions concerning health and
disease have enormous ethical implications.
2
Consider the emotion ‘despondency’ – should a despondent patient automatically qualify for treatment with
antidepressants? The decision medical practitioners make concerning treatment, may hinge on whether
‘despondency’ (simpliciter) is classified as a disease.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
10
In the following sections I outline three prominent conceptions of disease, looking
for that best suited to clinical medicine: those proposed by Rachel Cooper (2002), Peter
Schwartz (2007), and Jerome Wakefield (1992, 1999). I dismiss Cooper’s view on the basis of
clear cut counterexamples, but show that significant parts of both Wakefield and Schwartz’s
views are compelling. I conclude with a fourth proposal (which draws on both Wakefield and
Schwartz), providing an etiological theory of disease in clinical medicine, that is not, unlike
Wakefield’s etiological account, subject to the line-drawing problem.
2. The maximally value-laden conception – Rachel Cooper on disease
Cooper provides atripartite account of disease; that is, she proposes that the concept of
disease has three individually necessary, and together sufficient conditions.
(i) Disease is a bad thing to have;
(ii) We must consider the afflicted person to have been unlucky; and
(iii) The condition can potentially be medically treated. (see Cooper, 2002, 271)
The first criterion states that in order for a patient to have a disease, the condition
must be bad for that patient (as opposed to society at large). There are many examples of
one and the same condition being good for one individual, and bad for another. Having an
unusually fast metabolism, for example, is harmful to the bodybuilder (who wishes to
increase muscle mass), but beneficial to the model (who can, as a result, eat a balanced diet,
whilst maintaining a low body mass index (BMI)3). Cooper’s claim thus implies the
somewhat counterintuitive conclusion that the same condition can be a disease for one
person, and not for another. That said, her example of ‘sterility’ goes some way to justify
3
Here I assume that the model’s weight is not harmful to her in other respects, e.g. underweight to the extent
that her organs function poorly.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
11
this - some choose to have vasectomies, yet others, who desperately want children, are
sterile for reasons beyond their control. Only the latter group are diseased.
The second criterion states that an individual can be diseased only if they are
‘unlucky as judged by the uninformed layman, that is, roughly, worse off than the majority
of humans of the same sex and age’ (p. 276). She claims that this criterion helps us
understand why one can attribute states of health to those with disabilities, and to those
with genetic conditions such as Down Syndrome. Down Syndrome patients, she argues, are
healthy “in a certain sense”, just in case they do not have “some other” infection or injury.
(p. 276).
The third criterion, that the condition must be potentially medically treatable,
separates diseases from other harmful and unfortunate states; for example, being robbed at
gunpoint, losing one’s job, and so on.
Objections to Cooper’s model
Although the three conditions Cooper proposes do seem to apply to most diseases, the
theory is subject to a variety of counterexamples. First, it is questionable whether all three
necessary conditions are, in fact, necessary. I refer in particular to the second criterion, in
which she states that one must be “worse off than the majority of humans of the same sex
and age” in order to be diseased. Cooper suggests that “90-year-olds who can’t walk as far
as when they were younger are not diseased because we expect old people to become
increasingly frail” (p. 276). This seems fair. However, there is a clear correlation between
age and dementia; in an ageing population, it is worryingly possible that there will soon be
an age group in which dementia is the norm. Anyone who knows somebody with dementia
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
12
will confirm that the condition should always be deemed pathological, yet Cooper’s
conception implies that, for the very elderly, it might not be. This second criterion is at best
dubious, but there are more serious problems with the thesis. Cooper presents her view as a
set of necessary and sufficient conditions, yet there are numerous non-pathological states
that satisfy all three.
Cooper presents and refutes a counterexample of her own – that of unwanted
pregnancy. She argues that we consider this a counterexample, only because our intuitions
tend to lag behind the facts. Becoming pregnant can now (assuming one is taking
contraceptive measures) be ‘unlucky’, so unwanted pregnancy is a disease. The implication
is that, in time, we shall appreciate this.
Whether or not her response is satisfactory is questionable, but I shan’t dwell on this
example, since one can imagine far more robust counterexamples without difficulty. I
outline two possibilities below, but there are many more.
First, consider the pre-operative transsexual. A pre-operative transsexual feels he or
she was either ‘a man born in a woman’s body’, or vice versa (unlucky and bad), and one
can now have a sex change (medically treatable). A pre-operative transsexual is thus,
according to Cooper’s model, diseased. But of course, one does not consider the transsexual
to be diseased any more than the homosexual. He or she is perfectly physically able in every
respect, and the transsexual is not mentally ill4. Perhaps Cooper might respond that our
intuitions are lagging behind relatively recent developments in medicine, and that soon,
4
That is assuming, of course, that he or she is neither mentally nor physically ill in some other respect. e.g.
chest infection, schizophrenia, etc.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
13
now that the condition is medically treatable, we will consider the ‘condition’ of ‘preoperative transsexual’ a disease - but I suspect not.
Second, suppose Sam decides to get a tattoo reading ‘peace and love’ on her arm,
and, as is the fashion, decides to have it written in Chinese characters. This is explained to
the tattoo artist, whom Sam knows to be very skilled, honest and reliable. On this occasion,
however, the tattoo artist gets confused, and instead of ‘peace and love’, Sam finds herself
wandering around with “I hate China” decorating her upper arm (albeit in aesthetically
pleasing symbols). Given that Sam is about to go to China, this is definitely bad and unlucky,
but fortunately for her, also medically treatable. Nonetheless, it is not a disease. Cooper
might respond that the tattoo is, in fact, pathological, but she would be clutching at straws.
Let us assume that neither Sam nor the pre-operative transsexual is diseased. Why
do we take this to be the case? What strikes me is that neither Sam nor the transsexual has
damagedor poorly functioning physiological traits (furthermore, neither is mentally ill in any
respect). On the face of it, a plausible conception of disease in clinical medicine must
accommodate this intuition.
3. The pure statistical conception
In this section I consider whether a value-free conception of disease will resolve the
problems with Cooper’s tripartite account. Although the statistical thesis does rule out cases
like unfortunate tattoos and the pre-operative transsexual, the view ultimately fails (and for
countless reasons). However, the basic model plays a big role in both Schwartz’s account of
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
14
disease (discussed in section 4.), and in Boorse’s (1977)biostatistical theory (BST)5, so a
detailed exposition is warranted nonetheless.
Given the qualities expected of naturalist (value-free) views, and in particular, the
thought that scientific theories often involve statistical analysis6, it is unsurprising that some
have tried to differentiate between health and disease states using mathematical models.
The pure statistical model focuses on the quantifiable properties of an organism’s
physiological subsystems. The precise qualities to be measured differ depending on the
subsystem in question, of course, but the underlying method is (roughly) the same.
The statistical conception involves gathering data over large populations, and
applying formal methods to determine disease-status. Often (but not always), the measured
values of a characteristic (e.g. quantities of a hormone or cell count; sizes and shapes of
organs/tissues/cells; blood pressure, etc.) produce a normal distribution curve7. According
to the basic statistical model, some physiological subsystem is dysfunctioning, and diseasestatus is met, when the measured value of the characteristic under investigation is beyond
two standard deviations from the mean (Ananth 2008).
One might think this gives the thesis a normative quality, since the precise value at
which ‘the line is drawn’ (two standard deviations from the mean) seems somewhat
arbitrary; that is, nothing in nature ‘fixes’ this value. Perhaps surprisingly, however, this is
not particularly unscientific - at least insofar as it is not inconsistent with standard practice
5
See chapter 2 for a detailed discussion of the BST.
6
This is demonstrably true in the medical sciences. The results of epidemiological studies are almost always
grounded by statistical methods.
7
See figure 1 below.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
15
in other disciplines widely regarded as sciences. Central to testing significance in
psychological and epidemiological studies, for example, is the Null Hypothesis Significance
Test (NHST)8. When one attempts to establish a non-chancy relationship between two
variables, X and Y, the null hypothesis is that any correlation the data shows is accidental.
The NHST states that a result is significant only if P, the probability of the null hypothesis
being true (given the results of the study), is less than 0.05. This value, although not entirely
arbitrary (insofar as it would not be sensible to fix the value at 0.4 or 0.8 etc), was ultimately
chosen by R. A. Fisher early in the 20th century (Fisher 1925), not by nature.
The NHST is deemed a scientific method, yet the ‘line-drawing’ it involves is no less
arbitrary than that used in the ‘pure statistical method’ of picking out disease states. If it is
normal within the sciences to choose such arbitrary values, then there is no reason to think
doing so should fall outside the remit of naturalistic conceptions of disease; that is,
conceptions committed to employing only scientific methods.
Figure 1.shows a normal distribution curve plotting the distribution of values of a
property of some physiological subsystem (e.g. the quantity of a hormone produced by an
8
It is worth noting that this method has not gone unchallenged (Kirk 1996), but nonetheless it is still frequently
used.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
16
organ). The dotted lines represent standard deviations from the mean (those closest to the
y-axis represent the first standard deviation from the mean, and those furthest away
represent the second standard deviation from the mean). The mean, which lies on the yaxis, shows the average value of a trait across the population. The shaded sections thus
represent those members of the population outside of two standard deviations from the
mean value. Those members of the population falling under the shaded section, then, are
diseased with respect to whatever trait the curve represents; that is, the trait in question is
dysfunctioning.
One can see that the vast majority of the population’s members fall within two
standard deviations of the mean (only 4.55% are outside the two standard deviations). That
extremities indicate disease is, in some cases, very intuitive - if one’s body mass index (BMI)
is far lower or far higher than the statistical norm, then one is likely to have either an eating
disorder, or some other disease affecting one’s BMI. Furthermore, given that the model is
grounded by the quantification of physiological traits, in this way, it is clear that neither Sam
with her tattoo, nor the pre-operative transsexual, is diseased according to this model.
Nonetheless, the pure statistical model faces insurmountable objections.
Objections to the pure statistical model
The theory fails for a number of reasons. To mention just a few:
(i) Schwartz (2007) has noted that, at least intuitively, one can identify ‘healthy’ and
‘unhealthy’ populations – we would surely want to say that, given the (on
average) enormous improvements in nutrition and medical care, we (humankind)
are significantly healthier in the 21st century than we were in the 16th, but of
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
17
course, if one takes all and only the outliers (beyond two standard deviations) to
be diseased, exactly the same proportion of the population will be healthy at any
given time9;
(ii) There are many unusual physiological traits that are not diseases. For example, one
might have significantly more (or significantly fewer) benign moles than most
people, or one may have an extra toe, or a third nipple, or unusual hair colour.
(iii) Some unusual physiological traits are beneficial. For example, one may have an
exceptionally good immune system, or be athletically unusually able (Ananth
2008). One would not consider Mo Farah diseased because his physique is such
that he can run faster and further than most people;
(iv) There are many common diseases – diseases that far more than 4.55% of the
population suffer from (a recent study showed that 47.2% of adults in the United
States have periodontal disease (Eke et al 2012))
(v) On the face of it, clinical judgements depend on both the state descriptive and
normative content of patients’ condition (Ereshefski, 2009, 227). Consider a
woman in her early-thirties with small, non-cancerous, endometrial polyps
(noncancerous growths on the inside of the uterus). One possible symptom of
polyps is infertility. If a patient is rendered infertile by endometrial polyps, but is
otherwise symptomless (and will most likely remain so), then the statistical
model will deem her diseased - but whether she requires medical treatment will
depend primarily on whether she (ever) wants to get pregnant. If she does then
99
Kingma (2010) presents a similar argument, noting that in ‘harmful situations’, the vast majority of the
population can become diseased (e.g. universal exposure to a dangerous pathogen) – but this inference is
incompatible with the statistical theory, since the level of functional efficiency required to be healthy would
drop such that precisely the same proportion of the population were diseased.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
18
surgical treatment is necessary. If not, then no surgery is required. The clinical
decision thus depends both on the state description: the physiological state of
the patient; and a normative claim: whether the patient wishes to get pregnant.
In light of all these objections, one must reject the pure statistical view as a thesis in itself. In
the following section, however, I outline a more plausible concept of disease - one that uses
aspects of this statistical approach, whilst avoiding its primary flaws.
4. The frequency and negative consequences approach, and line-drawing problem
The problems with the pure statistical conception arise primarily from the same underlying
issue. According to the statistical model, the line that distinguishes healthy states from
diseased states is fixed at two standard deviations from the mean. In practice, were medical
records to be analysed in this way, one would discover that the location of the line differs
from one disease to another – it depends upon the proportion of the population suffering
from the negative consequences of the condition. The bad effects of tooth decay are
experienced by a large proportion of the population, so the line is considerably further to
the right than the statistical model would dictate. However, hypertrichosis (hair growth so
excessive that it is deemed pathological - sometimes known as werewolf syndrome) is rare.
On a curve representing quantities of body hair, since fewer than 2.5% of the population
suffer from hypertrichoses, the line would be drawn further than two standard deviations
away from the mean.
Schwartz (2007) has proposed a solution to the problem – the frequency and
negative consequences approach (FNC). According to this approach, the location of the line
varies depending on how serious and how common the negative consequences are.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
19
Figure 210:
The frequency and negative consequences approach
Schwartz introduces an additional dimension the statistical model - one representing the
negative consequences of physiological conditions (figure 2). The ‘negative consequences’
linein Schwartz’s model crosses the ‘prevalence in reference class’ line, where the negative
consequences experienced by the patient are deemed minimally ‘negative’ to be
pathological.
Schwartz’s model deals with most of the objections to the pure statistical theory: (i)
the healthy/unhealthy populations objection is bypassed, since a smaller proportion of
today’s population will satisfy the minimally negative consequences for most diseases, than
would have done in the 16th century; with respect to (ii) and (iii), none of the unusual
physiological traits that are not diseases (such as red hair, or being spectacularly fit) will
have the negative consequences necessary to be classified as pathological under the FNC;
10
With kind permission from The University of Chicago Press. Schwartz, P. H.Defining Dysfunction: Natural
Selection, Design, and Drawing a Line,Philosophy of Science,(July 2007): p.378, figure 2.a. ©The Philosophy of
Science Association. All rights reserved.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
20
and the common diseases problem is avoided, since the ‘negative consequences’ line will
cross the ‘prevalence in reference class’ line relatively close to the mean.
Although prima facie the addition of the ‘negative consequences’ dimension gives
the FNC a normative dimension, this is (according to Schwartz) not the case. He takes
negative consequences to be “effects that significantly diminish the [ability to] carry out an
activity that is generally standard in the species and has been for a long period of time”
(Schwartz, 2007, 379), and this, he claims, allows one to differentiate between functioning
and dysfunctioning ‘without problematic teleological assumptions or value-judgements’ (p.
384). If this is the case, then (unlike with Cooper’s heavily value-laden conception of
disease) two individuals cannot be in the same physiological state, and differ with respect to
their disease-status - if infertility due to polyps is a disease for one, it is a disease for all.
Objections to the frequency and negative consequences views
The FNC approach successfully dodges the line-drawing problem, but if the view is valuefree, as Schwartz intends11, then it is unsuitable for clinical decision-making. This is not a
valid criticism, since Schwartz does not present the FNC approach as the clinician’s concept
of disease, but as an analysis of ‘pathological condition’. I present the view here, rather than
in chapter 2, only because his response to the line drawing problem (or something very
much like it) is, I believe, necessary to accommodate the clinician’s concept of disease.
The aim of this chapter is to identify a conception of disease suitable for clinical
medicine, and one cannot make decisions about legal responsibility, eligibility for disability
allowance or sick pay, and so on, without making value-judgements. Although Cooper’s
11
Kingma has argued that Schwartz may have failed in this respect (2014, 596).
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
21
tripartite account of disease ultimately fails, the concept of disease she advocated was
‘personalised’, and this is a quality any successful conceptual analysis of disease in clinical
medicine must have. In the next section I put forward an account of disease as harmful
dysfunction (Wakefield 1992) - a view which, as the name suggests, includes both a
naturalist (value-free) and a normativist component.
5.1 The etiological account of natural function, and disease as harmful dysfunctions
The question of what a natural function is, and indeed, whether there are such things, has
been addressed extensively in the literature (Millikan 1984, 1989; Wright 1976; Ruse 1973;
Neander 1991). Accounts of natural function can, rather crudely, be divided into two: the
etiological and the non-etiological. Neither the statistical approach nor the FNC invoke any
kind of etiological notion into their analysis of disease. According to these conceptions,
whether one is in a disease state is at least partially, if not wholly (in the case of the pure
statistical account), based on the functional efficiency of one’s organs, tissues and cells,
relative to the functional efficiency of those of the population. Etiological accounts of
function, on the other hand, do not consider statistical notions relevant to natural function.
The natural function of a trait, for the etiological theorist, depends upon its evolutionary
history.
Both Wakefield’s conception of disease as harmful dysfunction, and the account of
disease in clinical medicine that I shall propose, require an etiological account of function.
The etiological account of function
It is generally accepted that the human heart was not intentionally designed to pump blood
around the body (that is, by some higher power/rational being), and that, after millions of
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
22
years of evolution, the best explanation for the presence of hearts is that they do, in fact,
perform this function (Wright 1976; Ruse 1973; Neander 1991; Wakefield 1992). That the
best explanation for the presence of hearts is the function they perform, naturally leads to
an account of proper (natural) function: The function of a trait is whatever effect that trait
was selected for.
To illustrate, consider Neander’s example of the opposable thumb:
‘…it is the function of your opposable thumb to assist in grasping objects, because it
is this which opposable thumbs contributed to the inclusive fitness of your
ancestors, and which caused the underlying genotype, of which opposable thumbs
are the phenotypic expression, to increase proportionally in the gene pool. In brief,
grasping objects was what the trait was selected for, and that is why it is the function
of your thumb to help you to grasp objects.’ (1991, p. 174)
In short, the etiological theory states that if a trait of a particular species has been
maintained during the evolutionary process, one should take the role it plays for that
species (qua the reason the trait has been maintained in the evolutionary process) to be its
function.
Objections to the etiological account of function
Following Ananth (2008), in this section I outline and refute two objections to the type of
etiological account defended by Neander (which Ananth labels ‘Etiological Functional
Naturalism’ (2008, p. 176)).
Objection (i)
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
23
‘…imagine a possible world in which the human lungs successfully [acclimatise] to an
atmosphere that has suddenly changed from being oxygen-rich to almost entirely
carbon dioxide-rich. The function of lungs before the atmospheric shift was to inhale
oxygen and exhale carbon dioxide, but in the new environment the lungs function to
inhale and exhale carbon dioxide. According to the etiological account, this new
activity of the lungs cannot be a genuine function of the lungs because it is not the
production of natural selection. The point of this counterfactual is that the
etiological account is forced to accept the counterintuitive view that many beneficial
acclimations that help keep an organism alive are not functions.’ (2008, p. 180)
Response to objection (i)
It may well be the case that the etiological account of function implies that some beneficial
acclimatisations are not functions, but whether or not this is overly problematic is not so
obvious. Arguments that appeal to intuition are questionable – if nothing else, because
intuitions often differ between parties. It is my intuition that in the example above, if the
inhaling and exhaling carbon dioxide-mechanism is not yet a part of a ‘natural selection
history’, then it is not (yet) a natural function of the lungs. Thatit is beneficial to survival and
reproduction (in Boorse’s sense) in the present circumstances, is not sufficient for ‘natural
function’ status12.
Objection (ii)
12
See Ananth, 2008, p 181 for further discussion.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
24
Human beings have a number of vestigial physiological subsystems and stimulus-response
mechanisms, the traditional example being the appendix13, but goose bumps, the
vomeronasal organ, and several others fall into this category, too. The etiological account of
function, however, seems to imply that these organs must still have a natural function14:
that which explains their existence.
Response to objection (ii)
One might think that Neander’s approach entails that the appendix has a function: to digest
plant-based foods with very little nutritional value. The appendix ceased to perform this
function some time ago (since we no longer eat these low-nutrition foods), so it no longer
contributes toward our natural goals, but nonetheless, helping with digestion might still be
deemed the natural function of the appendix.
Unfortunately, this is of little help to those wishing to incorporate an etiological
conception of natural function into an analysis of disease. If one grants that the appendix
still has a function, all vestigial organs are always diseased. The only option is to deny that
vestigial organs have a function, despite there being an evolutionary explanation for their
existence. This is not, however, as problematic as it first seems - one can add a further
condition that resolves the issue. I suggest one adapts the etiological account as follows:
(i)
A physiological subsystem has a natural function only if it contributes to the
organism’s natural goals.
13
Whether or not the appendix is vestigial is, in fact, questionable. Matsushita, Uchida and Okazaki (2007), for
example, conducted a study suggesting it plays a significant role in the pathogenesis of ulcerative colitis.
14
See Ananth, 2008, pp 181-182.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
25
(ii)
If a physiological subsystem has a natural function, it is that which explains its
existence in the organism.
One possible worry here, is that any view that appeals to dysfunction in its conceptual
analysis of disease will not be able to accommodate diseases of vestigial organs (such as
appendicitis). On closer inspection, however, this is not so problematic. Vestigial organs
cannot dysfunction, but they can cause other traits to dysfunction; in other words, although
a vestigial organ like the appendix can cause a disease, the organism is diseased in virtue
ofthe non-vestigial organ dysfunctioning.
5.2 Wakefield and diseases as harmful dysfunctions
Wakefield outlines his etiological account of disease as follows:
A condition is a disorder if and only if (a) the condition causes some harm or
deprivation of benefit to the person as judged by the standards of the person’s
culture (the value criterion), and (b) the condition results from the inability of some
internal mechanism to perform its natural function, wherein a natural function is an
effect that is part of the evolutionary explanation of the existence and structure of
the mechanism (the explanatory criterion). (1992, p. 384)
The value criterion permits this concept of disease to be used by the clinician, since it
accommodates the fact that two people with the same physiological condition can differ
with respect to disease status. The explanatory criterion ensures that the counterexamples
to Cooper’s conception are not applicable – Sam’s tattoo, although medically treatable, is
not the consequence of some internal mechanism failing to perform its natural function
(and so does not qualify as a disease), nor is the pre-operative transsexual’s - so far so good.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
26
Objections to Wakefield’s disease as harmful dysfunction
Wakefield’s approach, as it stands, fails to deal with Schwartz’s line-drawing problem. There
is a range of healthy levels of functional efficiency. One’s blood pressure is healthy just in
case one’s systolic blood pressure is below 140 and above 90, and one’s diastolic blood
pressure is below 90 and above 60. Wakefield claims his etiological approach can
accommodate this - the healthy levels of functional efficiency are distinguishable from the
pathological, since those in the healthy range were favoured by natural selection. However,
Schwartz writes:
[Wakefield’s] approach does not solve the line-drawing problem. The first difficulty is
that for acquired disorders, ones not caused genetically, natural selection does not
apply. For example, it is not the case that Mr. Smith’s [ejection fraction] of 20% is
favoured by natural selection or not favoured, since his condition is not hereditary.
(2007, p. 370)
On the face of it, Wakefield’s account of disease as harmful dysfunction is well-suited to the
clinician, since it permits medicine to be personalised. However, doctors also need to
distinguish the healthy from the pathological levels of functional efficiency, and as it stands,
Wakefield’s approach fails in this respect.
6. Disease as harmful function – ‘drawing the line’ on the etiological account of disease
Drawing on both Schwartz and Wakefield, in this section I propose new and feasible concept
of disease in clinical medicine – disease as harmful function.
Although Schwartz resolves the line-drawing problem, his view is not suitable for
clinical medicine. Wakefield’s approach, on the other hand, is prima facie a good candidate.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
27
However, Schwartz’s response to the line-drawing problem is not applicable to Wakefield’s
conception, and no alternative response is provided.Wakefield’s harmful dysfunction
accountthus includes the evaluative component necessary for the clinician, but it cannot
provide a naturalistic account of healthy functioning; that is to say, although an etiological
account of function can identify the natural function of a trait, it cannot identify the
associated range of ‘healthy’ functional efficiencies.
Schwartz’s argument is compelling, since there are many disorders for which
Wakefield’s proposed response fails. However, Schwartz’s response to Wakefield does not
rule out an etiological approach tout court.
As we have seen, the etiological account of natural function identifies the natural
function of a trait as the evolutionary explanation for that trait’s existence. Identifying the
naturalfunction, then, is not problematic for the etiological theorist. Providing a value-free
way of identifying thehealthy range of efficiencies would be, but fortunately, she doesn’t
have to.
Any plausible account of disease in clinical medicine is value laden, and indeed, the
first clause of Wakefield’s conception: “[a disease] causes some harm or deprivation of
benefit to the person as judged by the standards of the person’s culture (the value
criterion)” (1992, p. 384) is inherently so. The second clause, that “[the harm] results from
the inability of some internal mechanism to perform its natural function” (p. 384) is
problematic, for it forces Wakefield to provide an account of dysfunction. An account of
dysfunction, however, can be replaced with a list of trait-values (e.g. BMI values, blood
pressure values, quantities of insulin produced, etc.), plus the degree of harm/benefit
associated with those values (figure 3).
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
28
The etiological theorist can now offer an account of disease as harmful function. A
condition is a disorder if and only if “(a) the condition causes some harm or deprivation of
benefit to the person as judged by the standards of the person’s culture (the value
criterion), and” (b) the condition results from an internal mechanism performing its natural
function, at a harm-causing level of efficiency (that is, a level of efficiency that is bad for the
individual), “wherein a natural function is an effect that is part of the evolutionary
explanation of the existence and structure of the mechanism (the explanatory criterion)” (p.
384).
Figure 3:
Disease as harmful function
In figure 3 the range of healthy resting heart rates are represented by the grey arrow, and
all other resting heart rates are represented by the two black arrows. The function of the
heart is to pump blood around the body. Most of our hearts do pump blood around the
body within that healthy range, but according to the harmful functioning conception, that is
not what determines the healthy range. The range of healthy resting heart rates (where the
lines are drawn) is chosen by medicine. If a patient’s resting heart rate does not fall within
the healthy range; that is, if it is deemed to be at level of efficiency bad for the patient, then
according to the clinician, it is diseased.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
29
7. Chapter 1 conclusions
Cooper’s account of disease is heavily value-laden, and therefore equipped to deal with the
evaluative aspects of clinical medicine. However, this tripartite account is subject to many
counterexamples. One of the necessary conditions implies that dementia may not be a
disease for some, and there are numerous ‘conditions’ that satisfy all of her criteria, but are
clearly not diseases (e.g. having unfortunate tattoos).
The pure statistical view is value-free, and hence unfit for the clinician’s conception
of disease. Furthermore it is subject to counterexamples that render the view entirely
implausible as a stand-alone conception of disease. However, Schwartz incorporates the
statistical approach into his FNC approach, and although (as a value-free account) it is not
suitable for the clinician’s concept of disease, we shall see in chapter two that it is a
plausible candidate for the pathologist’s.
Wakefield’s disease as harmful dysfunction account fails because it cannot provide a
satisfactory response to the line-drawing problem; that is, it cannot draw the line between
where health ends, and dysfunction begins. ‘Disease as harmful function’, however, is valueladen to the extent required by the clinician; it ensures that diseases involve harm caused by
‘internal mechanisms’ functioning in an undesirable way; and this is not affected by its
inability to provide a naturalistic account of dysfunction, since dysfunction does not feature
in the view. Disease as harmful function, then, is a good fit for the clinician’s concept of
disease.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
30
Chapter 2. What is a Pathological Condition?
1.
Introduction
The target of chapter one was a conceptual analysis of disease for the clinician. Here I
analyse the concept of disease from a slightly different perspective. This chapter is a
conceptual analysis of the theoretical notion of disease, in which I ask: ‘what is a
pathological condition?’
Although the clinician’s concept of disease is clearly inherently value-laden, this is
not true of pathological conditions. The focus of this chapter is by far the most frequently
cited ‘naturalist’ conception of disease/pathological condition: that proposed and
developed by Boorse (1975; 1976; 1977; 1997; 2014). I ultimately argue the BST, as it
stands, has not been rescued from Guerrero’s (2010) Cambridge change objection.
Furthermore, I will show that by replacing the biostatistical conception of natural function,
with the etiological account of functionoutlined in chapter one, one can espouse a
conception of a pathological condition immune to all objections raised against the BST.
2. What is naturalism about disease?
Before outlining the BST, it is worth emphasising Boorse’s naturalist target. This is especially
important, since many of the objections to his thesis arise from confusion in this regard. For
example, Murphy writes:
The naturalist conception of disease (perhaps most clearly stated in Boorse 1975,
1997) is that the human body comprises organ systems that have natural functions
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
31
from which they can depart in many ways. Some of these departures from normal
functioning are harmless or beneficial, but others are not. The latter are ‘diseases’.
So to call something a disease involves both a claim about the abnormal functioning
of some bodily system and a judgement that the resulting abnormality is a bad one.
(Murphy, 2015, 2)
Although Murphy is right to state that the naturalist conception (or at least Boorse’s
naturalism) concerns departures from the normal functioning of physiological subsystems,
the claim that the naturalist concept of disease involves a normative judgement is entirely
misguided. Indeed, this is preciselywhat Boorse denies. He takes disease to be ‘a type of
internal state which impairs health’ (Boorse, 2014, 684) - but it is not value-laden, since
none of the ideas comprising the concept of pathological condition are value-laden15.
Boorse’s target is to ‘explain key concepts of biology and medicine’ (p. 695) – no more, no
less.
3.1Boorse’s naturalism
According to Boorse, physiological subsystems (organs, tissues, cells) have a natural goal: to
contribute to survival and reproduction. A subsystem is diseased, he says, when its
contribution to survival and reproduction is subnormal for the organism’s reference class16;
that is that relative to the biostatistical norm for the organism’s age, sex and species, the
subsystem makes a poor contribution to its natural goals.
Boorse outlines the BST as follows:
15
See objection (iii) to the BST
16
See below
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
32
1) The reference class is a natural class of organisms of uniform functional design,
specifically, an age group or sex of a species.
2) A normal function of a part or process within members of the reference class is a
statistically typical contribution by it to their individual survival and reproduction17.
3) A disease is a type of internal state which is either an impairment of normal
functional ability, i.e. a reduction of one or more functional abilities below typical
efficiency, or a limitation on functional ability caused by environmental agents.
4) Health is the absence of disease. (Boorse, 1997, p. 7-8)
The resultant picture resembles that of the pure statistical conception discussed in chapter
one, with the important exception that only subnormally functioning traits can be diseased:
Figure 418:
The biostatistical theory
17
Later revised to ‘survival or reproduction’, to accommodate reference classes that do not reproduce (Boorse,
2014)
18
With kind permission from The University of Chicago Press. Schwartz, P. H.Defining Dysfunction: Natural
Selection, Design, and Drawing a Line. Philosophy of Science(July 2007): p.272. figure 1 ©The Philosophy of
Science Association. All rights reserved.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
33
At this point, it is worth highlighting a few subtleties. First, Boorse’s reference classes
are not species simpliciter, but subclasses of species (determined by age groups and sex).
This prevents prepubescent children and postmenopausal women from being diseased due
to infertility; men’s testes, liver and adrenal glands being diseased for creating subnormal
(for the species) quantities of oestrogen (despite creating the mean level of oestrogen for
their sex), and many hundreds of similar examples.
Second, one possible interpretation of the second clause is: ‘A normal function of a
part or process within members of the reference class is a statistically typical contribution
by it to the survival of the individual organism, and to the ability of the organism to have
offspring”, since with respect to the former Boorse refers specifically to the organs and
processes of the individual organism, and there is little indication that he moves from
speaking of individual organisms to, say, their genes (or indeed to anything else). Boorse
later indicates, however, that ‘homosexuality’ may or may not be a disease - its status
depending on whether or not one of the kin-selection hypotheses is true (Boorse, 2014,
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
34
691). Homosexuality is undoubtedly detrimental to an organism’s propensity to produce
their own offspring, so one would think that the BST points toward it being classified as a
disease19; but if a kin-selection hypothesis is true, homosexuality is not detrimental to the
reproduction of the homosexual’s genes, since the individual’s genes are shared by his
sibling’s children (children whose survival is ex hypothesiaided by the homosexual). If the
disease status of homosexuality rests on the truth of one of the kin-selection hypotheses
being true, then Boorse’s natural goal of reproduction refers to the subsystems’
contribution to the reproduction of genes, and not to the individual organism’s ability to
have offspring. This is just as well, because Boorse intends the BST to be applicable across
the natural world (that is, not only for humans, but for non-human animals and plant life),
and there are many cases in which individual organisms cannot, by natural selection,
themselves have offspring. Worker bees and worker ants, for example, labour to ensure the
continuation and growth of their colony, but even when perfectly healthy, they cannot
reproduce – at least not in the way the ‘naïve’ reading of clause two suggests20.
This schema correctly identifies many diseases. One species-typical function of the
pancreas (more specifically, the beta cells within the pancreas), for example, is to produce
appropriate quantities of insulin to regulate blood-sugar levels, preventing hyperglycaemia.
This contributes to the organism’s natural goals, since it helps prevent myocardial infarction,
kidney failure, and stroke (all of which are detrimental to individual survival and
19
Note that, according to Boorse, the concept of disease is notvalue-laden, so those who believe the
biostatistical theory in some way degrades homosexuals misunderstand the thesis.
20
This is nicely consistent with contemporary thought in biology. See Dworkins 1976; Llyod 2007; Smith 1998
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
35
reproduction!). According to Boorse, if the pancreas is producing subnormal quantities of
insulin, the organ is diseased, and of course, medicine agrees21.
One can also see how this analysis does not fall foul of at least some of the
objections to the ‘pure’ statistical conception. The abnormal qualities of Farah (or at least
those contributing to his athleticism) do not impair the functional efficiency of his biological
subsystems. Indeed, the functional efficiency of his muscles and lungs is clearly significantly
higher than most, since in a situation where running is necessary (say, from prey), their
contribution towards his natural goals is unusually good. Given that Farah’s lungs are not
functioning subnormally, they are not diseased. A similar response applies to the ‘third
nipple’, and ‘unusually few benign moles’ examples. They do not reduce functionality with
respect to the organism’s natural goals, and thus do not count as diseases, either.
Furthermore, the second disjunct of the third clause accommodates (at least some) cases
where disease is the statistical norm, avoiding the usual counterexamples like gum disease
(Boorse 1997).
3.2 Objections to Boorse’s naturalism
In the following sections I present a number of existing objections to Boorse’s naturalism,
some more compelling than others. Most, however, can be responded to without too many
concessions.
Objection (i) The thesis is both too strong, and too weak
21
To have a pancreas producing subnormal quantities of insulin, is to have one of a variety of forms of
diabetes.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
36
Boorse’s conception prima facie looks both too strong, and too weak. Too strong, since as
Joseph Margolis has pointed out:
[It] is quite possible to imagine a set of circumstances in which eyes would lose their
function and yet not be diseased. For example, imagine that, because of terrestrial
pollution, the human race adopts, and adapts to, a life maintained at a submarine
level unpenetrated by water. (Margolis, 1976, 247).
Similarly,
one
can
choose
to
make
one’s
biological
subsystems
perform
subnormally/malfunction/dysfunction, and subsequently choose to restore their functional
efficiency e.g. when one uses earplugs, wears a blindfold, or dives underwater (Flew 1973).
Yet clearly one is not diseased when the visual impairment is intentionally caused by an
artificial, external item, which can be removed at will.
The thesis is too weak, since one can be diseased without subsystems
malfunctioning. For example, a patient may have had multiple epileptic seizures, but thanks
to effective medication be seizure-free for several years. It would be odd to claim this
patient does not have a disease (and one requiring treatment, at that), when, if the patient
stopped taking the medication, the seizures would return.
Response (i)
These objections can be dealt with fairly easily. First, the strange scenario Margolis outlines
is unproblematic for Boorse, since the species-typical contribution of the eyes to individual
humans’ reproduction and survival would be negligible in those circumstances; second, to
accommodate blindfolds and earplugs, one can include a clause not permitting cases where
one can choose to restore one’s natural functions to a healthy state - one can shift the focus
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
37
from actual subnormal functioning, to dispositions to dysfunction. In fact, Boorse intends
the biostatistical theory to be read, in this way. He quite clearly states in his 2014:
[In my 1977 I required] “functional readiness”, not just current function, and noted
that “biological functions are usually performed on appropriate occasions, not
continuously. (p. 685)
If one has healthy eyes but is blindfolded, one retains the disposition to see when the
blindfold is removed; and even whilst taking the epilepsy medication, an epileptic is
disposed to have seizures under normal conditions (that is, when no medication is being
taken).
Objection (ii) There are many possible reference classes
Boorse’s reference class criterion is vital to the success of the biostatistical theory, but
Cooper has argued that this clause is problematic, not because separating into reference
classes is unnecessary, but because separating into sex, age, and species-type is insufficient.
Empirical evidence, she argues, suggests that if one were to adopt an account of disease of
this nature, considerably more reference classes than Boorse proposes are necessary.
What’s normal for an organism depends not only on species, sex and age, but also on
a host of other factors. Masai are naturally sensitive to growth hormone, pygmies
are not [, and athletes] normally have a lower heart rate than other people. (Cooper,
2002, 266).
To accommodate the lower levels of oxygen in the air, those who live at high altitudes (on
average) have a higher red blood cell count than those who live at lower altitudes; so what
counts as a subnormal red blood cell count should, it seems, be lower for those living at
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
38
lower altitudes. If one grants this prima facie intuitive conclusion, then the reference classes
must include more categories than just those of age, sex and species. Cooper argues one
must include (at least) race, environment, and training, as well22.
Of course, introducing additional reference classes reduces the average populationsize. If the classes become too small, then the thesisbecomes implausible, since the smaller
the population-size, the less likely it is that the distribution curve will cohere with the
medical consensus on which physiological states are pathological. Cooper even suggests
that the population of some reference classes could be just one, in which case, every
member of that population (i.e. the only member) must be healthy in every respect, for she,
alone, determines the mean levels of efficiency.
Response (ii)
Rather than attacking the ‘many references classes’ objection from an ‘in principle’
perspective, Boorse (2014, pp. 702-705) takes each of the classes proposed by Cooper, and
disputes them individually.
Cooper suggests that race must be a reference class - but Boorse has already argued
in his 1997 that it may be necessary to add race to his list. Adding race would not be overly
problematic, since race (alone) would not reduce the size of the reference classes
significantly enough for Cooper’s objection to hold any weight.
With respect to environmental factors, the response to objection (i) is equally
applicable; if one is healthy, one’s physiological subsystems are disposed to perform their
22
‘Training’, since marathon runners and cyclists etc. will often have slower heart rates than relatively
sedentary people.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
39
functions with (at least) species-typical efficiency in typical circumstances. Those who have
adapted to live at high altitudes have a higher red blood cell count than those living at low
altitudes, but the circulatory systems of healthy individuals at both high and low altitudes,
are disposed to perform their function with at least normal efficiency in typical
circumstances; that is, in typical circumstances, those individuals would have at least
species-typical red blood cell counts.
Finally, Boorse dismisses the physical training examples as “a joke”. It is of course
true that the typical heart rate of a professional cyclist is lower than that of a non-cyclist;
but a professional cyclist with a heart rate typical for a non-cyclist does not (automatically)
have a pathological condition. The species-typical heart rate is medically normal, so
assuming the cyclist’s heart is not disposed to function subnormally in typically
circumstances, the cyclist is not diseased. (p. 703).
By addressing each of Cooper’s proposed references classes in turn, and showing
why none of them (with the possible but not unproblematic exception of ‘race’23) should be
included in his schema, Boorse has undoubtedly diffused Cooper’s argument. It has not
been categorically proven that further references classes are unnecessary, of course.
Perhaps classes other than those proposed by Cooper are required, and perhaps better
counterexamples to the existing thesis exist – but this looks unlikely, and until they are
provided, the BST cannot be doubted on these grounds.
Objection (iii) Thebiostatistical view is value-laden
23
As stated, Boorse’s hypothesis can accommodate ‘race’ as a reference class, anyway.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
40
As we have seen, to avoid counterexamples like ‘children must be fertile to be healthy’,
Boorse introduces the reference classes of age, sex, and species-type. But what if the
references classes had been chosen differently? By choosing age, sex and species-type,
rather than ‘having Down’s syndrome’ and ‘long-term alcohol abuse’, Boorse is able to claim
that Down’s syndrome and liver cirrhosis are diseases (Kingma 2007); however, the BST
strategy would be forced to conclude otherwise were the references classes different
(DeVito 2000). So what justifies Boorse’s particular selection of reference classes? Although
it clearly makes more sense to choose the classes of age, sex and species-type than those of
‘long-term alcohol abusers’ (partly, perhaps, precisely because the results cohere better
with those states we deem pathological), the mere act of choosing reference classes(it is
argued) involves making an evaluative judgement.
The biostatistical theory also implies that if the function of some subsystem pertains
to anything other than the goals of survival and reproduction, then its efficiency is irrelevant
when assessing disease states. This seems wrong. Human beings have goals that reach far
beyond those of survival and reproduction. One would imagine, for example, that
successfully fulfilling one’s goal of writing good philosophy is far less conducive to
reproduction than successfully fulfilling one’s goal of being a rock star, but writing good
philosophy and becoming a rock star are equally legitimate goals, regardless of the
respective conjugal implications. Indeed, one may actively desire not to reproduce. Should
one deem all those who have had vasectomies to be diseased? The ‘common sense’
conclusion is surely “no” - but those who have chosen to have vasectomies unquestionably
have a subnormal ability to reproduce.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
41
If one does not think all those with vasectomies are diseased, then either the list
of/criteria for ‘natural’ goals requires revision, or one needs to abandon the notion that
natural goals feature in the concept of disease altogether, perhaps in favour of more
general value judgements. The latter, of course, is a normativist strategy through and
through. Either way, whether one chooses ‘survival and reproduction’, or alternative goals
pertaining to more general wellbeing, purely in choosing the goals it seems one is making a
value judgement. In short: one makes judgements both in choosing the ‘natural’ goals, and
in choosing the reference classes, and a naturalist account involving evaluative judgements
is no naturalist account at all.
Response (iii)
Objection (iii) is a popular one, but it is entirely misguided. Boorse sets out to provide a
conceptual analysis of disease consistent with scientific method, using only the language of
science - his aim being ‘in some sense an attempt at a “lexical definition” (1977, 551) of
scientific terms’ (2014, 713).
The key point is this: for a concept to be value-laden, it must include value-laden
terms or ideas. The biostatistical theory does not. Neither survival nor reproduction nor age
nor sex nor species are value-laden terms. One might push that survival and reproduction
form part of the concept only because they are the natural goals, and ‘goal’ is a value-laden
concept - but that won’t help. Boorse’s four-part schema describes functions only in terms
of the contribution of subsystems to survival and reproduction (there is no mention of goals,
here) - neither the idea of survival, nor the idea of reproduction, is value-laden.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
42
That the component parts of the disease-concept are what they are, may involve
some kind of choice, but the concept of disease ‘already exists as a target [in pathology]… So
the only way to run an argument of this type is to claim that [pathology] – not the [BST] –
has chosen one of many possible health concepts.24’ (2014, 693). Medicine chose the
content of the concept of disease, but only in the same sense that biology chose the criteria
for being a mammal. That does not make ‘mammal’ a value-laden concept!
Boorse goes on to write:
None [of my critics] gives sense to the idea of a value-based “choice of the health
concept” by [pathology]. The obvious way to do so fails. That is to assume from the
outset that ‘disease’ is an evaluative concept, meaning something like “undesirable
condition” or “condition deserving medical treatment.” Then [pathology] will choose
what falls under this description – i.e. make value judgements about what conditions
are undesirable or need treatment. But, so clarified, the argument has two fatal
defects. First, it is circular, since it assumes its conclusion, that ‘health’ is valueladen. Second, it ignores one of the most basic features of medical usage of
‘disease’: that disease and medically treatable condition do not coincide. (p. 693)
Boorse must surely be right. Consider the case of the vasectomy. One may argue
over whether or not being infertile is always a state of disease (whether naturally occurring
or self-inflicted), but surely one would never consider being fertile a disease! Nevertheless,
we can, and do, medically ‘treat’ fertility with vasectomies and other contraceptive methods
24
Where I have written ‘pathology’, Boorse writes ‘medicine’. This replacement is useful to avoid confusion
with other contexts in which the term ‘disease’ is used, and warranted, since Boorse explicitly states that he is
analysing the concept of ‘pathological condition’. I shall not replace ‘medicine’ with ‘pathology’ throughout
this chapter, but the same substitution can usually be made.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
43
when high fertility is an undesirable physiological state. ‘Disease’ and ‘medically treatable
condition’ are altogether different concepts.
For Boorse, it is determined empirically that the goals and reference classes specified
arethe right goals and reference classes. We have already seen that there are several
candidate reference classes. I have spoken of age, sex, species-type, race, training,
environment, long-term alcoholics, and people with Down syndrome, but of course there
are indefinitely more possibilities. For the naturalist, though, some of these are the right
references classes, and the rest are not. In medicine, whether or not Dis a disease is an
empirical matter of fact. For mental disorders, for example, one need only look at The
Diagnostic and Statistical Manual of Mental Disorders (DSM-5). If the right goals and
reference classes are those that correctly identify diseases (which seems to be the case, if
one is searching for a lexical definition), then discovering the right goals and reference
classes is purely a matter of empirical enquiry, with no normative component25.
So why does the normativist think the naturalist has a problem, here? Here is one
possibility:the theory ‘Constructivism’ in the philosophy of medicine (as presented by
Murphy 2015), and the fact that the concept of disease has been, in some sense,
‘constructed’ (by medicine), have been conflated. Boorse’s naturalist view is constructed,
but only in the sense that the concept of fish has been constructed (after all, biology could
have chosen ‘having fins’ to be sufficient for being a fish). But that there is a choice
25
It is worth noting that Boorse does not believe that those in the medical profession alwayshave it right, so his
thesis includes a small proviso: ‘that scientists are sometimes confused, inconsistent, or (as with fever)
empirically wrong about their subject’ (Boorse, 2014, 713). The biostatistical theory, then, is not merely a
descriptive thesis, but a prescriptive one – it tells us which physiological conditions should be deemed
pathological.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
44
component to the concept of disease does not imply that the concept it is value-laden. It
does not imply that it is ‘Constructivist’ in Murphy’s sense below.
The key [C]onstructivist contention is that there is no natural, objectively definable
set of human malfunctions that cause disease. Rather, constructivists assert that to
call a condition a disease is to make a judgment that someone in that condition is
undergoing a specific kind of harm that we explain in terms of bodily processes.
(Murphy, 2015, 3)
Murphy’s Constructivism clearly involves a normative component, but the BST does not
resemble Constructivism at all – the BST makes no reference whatsoever to patients
undergoing harm. Although there is of course a ‘choice’ component to Boorse’s concept, as
there is with most concepts in the sciences (and indeed more generally), that concept
involves no value-laden ideas, and is hence not value-laden26.
Objection (iv) The Cambridge change objection
It is not uncommon in the medical sciences for conclusions to be reached by comparing the
qualities of one reference class to another, or of one individual to another; the results of
these contrastive studies depend both on the state of an individual (or of a cohort), and on
the state of the reference class. Since populations are subject to change, both with respect
to the members of the population, and with respect to the properties of the members, what
is statistically typical is also subject to change; according to BST, then,health status is subject
to change without any physiological change in the individual.
26
For a fuller and very cogent account of how constructivism, so understood, can dovetail with the BST, see
Kingma 2012.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
45
Suppose the BMI of >30 indicates disease. According to the BST, this fact is grounded
by what is species-typical. The BMI level indicating disease will thus increase, if the average
BMI of the population increases. Now suppose you have a BMIof 30, and currently qualify as
diseased under the BST. Should you believe that your health improves as the speciestypicalBMI increases to 30, even though your own weight remains constant? Presumably
not, since you are equally susceptible to myocardial infarction, diabetes, and so on. A
change in disease-status without a change in intrinsic physiological qualities is known as a
‘Cambridge change’ (Geach 1972), and the thought that health status changes can be
Cambridge changes, is surely absurd (Guerrero 2010).
Response (iv)
Boorse provides a three part response:
(1) the theoretical possibility of a Cambridge change in someone’s health status
seems to be entailed by any view of health as normality, an idea basic to scientific
medicine. But (2) it is far more difficult than Guerrero thinks for such a change to
occur, and (3) the only realistic ways in which one could occur are of no importance
to medical practice. (2014, p. 715)
These responses are convincing if one accepts that health as normality is basic to medicine.
Given that Boorse’s target is a lexical definition of scientific terms, this, I suppose, is an
empirical matter. However, even if it is true, one must question whether the concept of
health might lose its connection to normality if any Cambridge change situations actually
occurred. If one is using the Cambridge change objection to question conceptions of health
as normality, however, response (1) is question-begging.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
46
With respect to (2), prima facie Cambridge change situations are not as difficult to
imagine as it might seem – the sudden outbreak of bubonic plague in the middle ages, for
example, would have a significant effect on the average functional efficiency of many traits,
and of course, that event was of significant importance to medical practice. Boorse has a
response to this, however. The BST does not, he says, base normal health on the speciestypical levels of functional efficiency at particular ‘moments’, but over temporally extended
‘time-slices’ (Boorse, 2002, 99). This means that a quick change in the mean efficiency of
some subsystem will not significantly affect the health status of individuals, since those
individuals at the present moment form only a small part of the population over the timeslice (in its entirety).
Suppose a meteor hits the Earth and kills all human life with the exception of one
individual, who merely loses an eye. Were one to ask: ‘How many functioning eyes does a
healthy human have?’, given Boorse’s additional criterion, the answer would be two, since
for a long period prior to this unfortunate celestial encounter (that is, during the rest of the
time slice), the mean number of functioning eyes for a human being was two. The fact that,
at this precise moment, the mean number of functioning eyes is one, does not affect this in
any significant way. The same principle applies to rapid spreads of infection, and so on.
This response, although initially appealing, is at best incomplete. The ‘time-slice’
Boorse appeals to must be specified (or at the very least, more needs to be said about it),
for very lengthy time-slices often produce as many counterintuitive consequences as the
‘single-moment’ time slices assumed by Guerrero. Most of us are happy to say that the
average health of human beings, at least in western countries, has improved over the last
500 years or so; not least because improvements in nutrition, hygiene, and developments in
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
47
medical care have significantly increased both life expectancy, and quality of life. On the
face of it this is supported by Boorse’s hypothesis, since the average species-typical
contribution of our physiological subsystems to our natural goals is higher than it was in
1515 CE. However, if one takes the time-slice factor into consideration, and suppose
Boorse’s time-slice covers 500 years, many people now considered to be malnourished and
diseased would be classified as healthy, since what is species-typical over the time-slice
partly depends on the (relatively) poor physiological conditions of those living in the 16th
century. Boorse does suggest that any period shorter than ‘lifetime or two’ is too short for
the necessary time-slice27, so ‘one can only conclude that [Boorse thinks] the time slice is
between two lifetimes and a millennium’ (Giroux, 2015, 185). In light of the above, it should
probably be closer to two lifetimes.
Élodie Giroux (2015), however, has recently (and reasonably) claimed that the
concept of health should cohere with developments in epidemiology. After all,
epidemiological research is of central importance to the prevention, diagnosis and
treatment of disease28. She argues that Boorse’s suggestion of between two lifetimes and a
millennium is far too long for the time-slice, since it could not accommodate
epidemiological studies:
‘…theoretical medicine and physiology cannot ignore epidemiological changes in
human morbidity and in human longevity… Thus, far from supporting the idea of a
stable and natural human functional design, descriptive epidemiology reveals the
importance of historical evolution of health phenomena at the population level and
27
See Boorse (2002)
28
Much more on this later.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
48
brings to light a crucial issue: the relevant time slice extension of a reference class…
it seems to me that a shorter population time-slice to which our theoretical health
judgments are relative would better fit contemporary medicine.’ (Giroux, 2015, 186)
If the time slice is short, then Guerrero’s Cambridge changes will be rife; if it is centuries,
then the average functionality will be so low, that far too many present people will be
deemed healthy; and if it lies somewhere in the middle, say, two lifetimes, then it is
incompatible with current assumptions and progress in epidemiology and clinical medicine.
A convincing response to the Cambridge change objection may appear at some point, but
this objectionis certainly a threat to the BST.
4. The frequency and negative consequences approach revisited
Schwartz’s FNC approach, like the BST, is offered as a value-free conception of pathological
condition. Schwartz also takes disease to be a matter of subnormal functional efficiency,
and although most of the objections raised against Boorse would also be raised against FNC,
the FNC approach is able to use the same responses. But Schwartz can go one better.
We saw that in order to resolve the line-drawing problem, Schwartz suggests that an
extra dimension to the BST should be added – that of ‘negative consequences’;
whereinnegative consequences are “effects that significantly diminish the ability of a part or
process in the organism… to carry out an activity that is generally standard in the species..”
(2007, p. 379))29. This additional dimension allows the FNC approach to combat the
common diseases problem, but it also resolves the Cambridge change objection. The
Cambridge change objection states that the levels of functional efficiency necessary to be
29
See chapter one section 4
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
49
healthy, rises and falls with the species-typical functional efficiencies – thus one can be
healthy one day, and diseased the next, despite no changes in the functional efficiency of
one’s own traits. This does not apply to the FNC approach, however, since the negative
consequences condition moves ‘the line’ as the average levels of functional efficiency rise
and fall. As the average level drops, the line moves to the right such that a higher proportion
of the population are diseased; and as the average functional efficiency rises, the line moves
to the left, such that a lower proportion of the population are diseased.
The FNC approach is promising, but on the face of it not flawless. Kingma questions
the naturalism of the negative consequences factor, pointing out that “…sometimes, where
the situation demands it, “the ability to carry out a standard activity” needs to be
suppressed or impaired” (Kingma, 2014, 596)30 – one cannot breathe and swallow at the
same time, but this is not pathological. A value judgement needs to be made, then, to
determine which abilities can be suppressed without affecting health status. The FNC may
or may not be a purely naturalist view, but either way, it does deal with the Cambridge
change objection that the BST looks to struggle with.
The target of this chapter is a plausible conceptual analysis of pathological condition
– I do not assume that this must be entirely value-free. Both Boorse and Schwartz claim
their views are value-free, since neither (they claim) contains value-laden concepts; that
said, if Schwartz’s negative consequences condition is value-laden, this only shows Schwartz
to be mistaken about that. It does not imply that the FNC is implausible as a theoretical
conception of disease - the FNC would not permit, for example, two individuals with the
same set of trait functional efficiencies, to differ with respect to their health status.
30
It is worth noting that Kingma does not consider this a destructive objection, referring the reader to
Hausman (2012). It is not within the scope of this chapter to investigate this further, however.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
50
5.1 The etiological theory of pathological condition
The FNC approach is plausible if one prefers non-etiological conceptions of natural function,
but otherwise not. In this final section I propose a further naturalist account of pathological
condition; this view, like the BST, states that diseases are impairments to the biological
functions that contribute to survival and reproduction. The view I propose - the etiological
theory of pathological condition - draws heavily on Boorse and Schwartz, but rejects the
strong link between theoretical health and biostatistical normality. Instead, the crucial
second clause is replaced with an etiological account of function.
Below is the ‘etiological theory of pathological condition’ (ETPC). Note first that the
effect for which a physiological subsystem was chosen differs between reference classes –
natural selection chose that healthy prepubescent boys cannot reproduce, and natural
selection chose that adult men can. Clause 2 refers to the ‘reference class relative effect for
which that trait was selected’. This covers the need for the natural function(s) of some traits
to differ between reference classes.
1) The reference class is a natural class of organisms of uniform functional design,
specifically, an age group or sex of a species.
2) The natural function of a trait within members of the reference class is the reference
class relative effect for which that trait was selected, unless it fails to contribute to
individual survival, or reproduction of the organism’s genes. In which case that trait
has no function.
3) A disease is a type of internal state which is (a) either an impairment of a natural
function; i.e., a reduction of functional ability, or a limitation on functional ability
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
51
caused by environmental agents; and (b) is deemed by medicine to have sufficient
negative consequences (to a level chosen by the pathologist) to be pathological,
where negative consequences are “effects that significantly diminish the ability of a
part or process in the organism… to carry out an activity that is generally standard in
the species”.
Now let us return to the Cambridge change objection. ETPC is not threatened by
Cambridge changes. It draws on our best biological theories to produce an account of
function for biological-types, where the natural function of a trait is the reference class
relative effect for which that trait was selected. ETPC identifies disease wherever a
physiological state impairs a natural function31, such that it has at least the minimally
negative consequences for the organism, as chosen by the pathologist. This makes no
reference to biostatistical normality – functions are defined etiologically, and health status is
determined only by internal physiological properties, and whatever pathology considers a
sufficiently negative consequence (in Schwartz’s sense) for a condition to count as a
pathological.
5.2 Objections to the etiological theory of pathological condition
Objection (i)
Given that the ability of the reproductive organs to aid reproduction is the explanation (or at
least one of the explanations) for their existence, ETPC must account for the fact that the
inability of young children to reproduce in not pathological.
Response (i)
31
Or there is a limitation on functional ability caused by environmental agents.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
52
We have not evolved to reproduce as young children32. Clause two specifies that functions
are reference class relative, so any objection of this kind can be dismissed.
Objection (ii)
ETPC fails to distinguish between old age impairments, and normal impairments.
Response (ii)
Boorse suggests that, given that different reference classes provide different biostatistical
norms, some conditions are diseases for the elderly but not for young adults. In many
instances, my conception does not ‘accommodate’ this, but I consider it an advantage of the
etiological theory of disease. Surely Alzheimer’s disease is a disease for the elderly as well as
for the young (even if most elderly people were to suffer from the condition, which is an
increasing worry in our aging population); surely a pathologist can identify coronary heart
disease in the elderly as much as she can in the young... The etiological conception of
disease is grounded by evolutionary theory, so assuming our brains and hearts have not
evolved to dysfunction, the pathological conditions of young adults and the elderly coincide.
That is not to say these conditions should be classified as identical - the same impairment to
a natural function might have to be classified and treated differently depending on the age
of the patient. Early onset Alzheimer’s and early onset Parkinson’s, for example, are treated
as different medical conditions to the Alzheimer’s and Parkinson’s diseases suffered by the
32
In fact, the etiological theory of disease deems fertile children to have a pathological condition. The early
onset of puberty, or ‘precocious puberty’, is a recognised pathological condition. Not only is fertility unsuitable
for children because they are not sufficiently (mentally) mature, but precocious puberty causes the early
maturation of bones, and consequently the early cessation of linear bone growth.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
53
elderly. Nonetheless, both are pathological conditions, and the naturalist only sets out to
differentiate between health and disease states – it is not a means of classifying diseases33.
Objection (iii)
One might reiterate Cooper’s objection to Boorse’s theory, that there are more reference
classes than simply age, sex, and species-type (see objection (ii) in 3.2 of this chapter).
Response (iii)
Applying a Neander-style etiological account of function to a theory of disease also deals
with the ‘too many reference classes’ objection. If one takes the natural function of a trait to
be the reference class relative explanation for its existence, and a disease to be the
malfunctioning of that trait, what is statistically typical for a reference class becomes
irrelevant. It does not matter whether there is only one member of a reference class, nor
does it matter how many reference classes there are34.
Objection (iv)
ETPC is not value-free. First, an etiological approach to natural function has a normative
component – it tells us what physiological subsystems are supposed to do, not just what
they, in fact, do. Second, that pathologists choose where to draw the line between the
healthy and the pathological implies that the theory is value-laden. Given that the approach
is value-laden, it is not a suitable candidate for the theoretical notion of disease.
33
Of course, this argument flounders if physiological subsystems have evolved to dysfunction, but given the
aims of medicine, medicine does not (nor should it) consider this possibility relevant to either clinical practice
or pathology.
34
For the record, I am happy with Boorse’s response that medicine (pathology) has chosen the reference
classes, anyway.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
54
Response (iv)
Given that the ETPC rests on an etiological account of function, the fact that the view has
normative components is undeniable. However, Karen Neander (1991) embraces the
normative and teleological qualities of her etiological account of natural function, asserting
that “the biological notion of ‘proper function’ can be both teleological and scientifically
respectable” (1991, 454). ETPC contains no qualities that are not ‘scientifically respectable’;
further, as I argued at the end of section 4of this chapter, that there is a normative
component to a theoretical conception of disease, does not render it implausible.
6. Chapter 2 conclusions
In this chapter I have detailed the most commonly discussed naturalist conception of
disease, the BST. Many have objected to the BST, the most common objection being that
the view is not, as it claims to be, value-free – but Boorse is right to say that this objection is
misguided. For a concept to be value-free, it need only contain no value-laden ideas, and
the BST does not. The BST is, however, susceptible to Guerrero’s Cambridge change
objection.
The FNC approach is immune to Cambridge changes. Schwartz introduces a
component that takes into consideration the negative consequences of a condition, so as
populations become healthier (or less healthy), the line which demarcates the healthy from
the diseased portions of the population shifts accordingly. Schwartz’s approach, if one is
willing to accept a non-etiological conception of natural function, is a plausible theoretical
conception of disease.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
55
In the final section I proposed the etiological theory of pathological conditions.
According to this view, diseases are impairments to natural functions with negative
consequences (to a degree determined by pathology), where the natural function of a trait
is that which explains its existence in the organism. This differs fundamentally from the
etiological conception of disease in clinical medicine, since no two individuals could be in
the same physiological state, but differ with respect to their disease status. The view is
value-laden, but only in a perfectly scientifically respectable way – the only value-laden
qualities of the thesis are grounded by evolutionary theory, and the informed choices
pathologists make, concerning which impairments to biological function have sufficiently
negative consequences, to be deemed pathological. It is therefore a viable alternative for
those who prefer etiological accounts of function to their non-etiological counterparts.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
56
Part One Conclusions
Given that clinicians are concerned primarily with resolving the complaints of their patients,
it is unsurprising that the conclusion reached in chapter one involved a large normative
component – if a condition is not bad for an individual, then it is not a disease for her; the
very same condition, may, however, be a disease for someone to whom it is harmful.
Note, however, that the concept of ‘disease’ and the concept of ‘harmful, medically
treatable condition’ differ significantly (this is true for both the clinician and for the
pathologist) – they must do, since many cases of the latter are not diseases. To
accommodate this, a conceptual analysis of disease in clinical medicine must include a
concept of proper function. I presented a number of options, but concluded that the most
viable concept of disease for the clinician is one of ‘harmful functioning’. This permits an
etiological view of natural function; accounts for the clinician’s need to make evaluative
choices; and successfully distinguishes between diseases, and those harmful, treatable
conditions that do not fall into this category (such as unfortunate tattoos).
In chapter two I focussed primarily on Boorse’s BST, but I ultimately endorse another
etiological account – the ETPC. This account of pathological condition has a normative
component, but the normative component does not concern the harm caused to the
individual patient, in Cooper’s (2002) sense; it is normativebecause it is grounded by an
etiological conception of natural function. Given that the presupposed etiological account of
function is scientifically respectable, and that the view does not factor in whether or not a
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
57
condition is harmful to the individual (in Cooper’s sense), the ETPC being value-laden is
unproblematic.
We can see that the clinician’s concept of disease is not coextensive with that used
in pathology. For the clinician, if a condition is to be considered a disease, it must cause the
individual patient harm; that is, it must be bad for the patient. For the pathologist, on the
other hand, if a condition is a disease for one, it is a disease for all. The concept of disease,
then, is context dependent.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
58
Part Two
Disease and Causation
Chapter 3. Concepts of Causation in the Philosophy of Disease
1.Introduction
Causation is fundamentally important to all areas of medicine: the GP diagnoses the cause
of her patient’s discomfort, prescribing drugs, rest, and therapy, based on her knowledge of
the determinants of disease, and of treatments (interventions); the pathologist sometimes
looks to identify a cause of death; and the epidemiologist conducts studies to discover the
determinants of disease, and both preventative, and curative interventions.
In this and the next chapter I discuss two related, but distinct topics concerning
causation in medicine. Chapter 4 looks at causal inference in public health. There I focus on
the potential outcomes approach, a popular theoretical framework to effect measurement
in epidemiology. In this chapter I outline three conceptual analyses of causation in
contemporary analytic philosophy, considering their actual, and potential applications in
both evidence-based practice (EBP), and medical research. As with the concept of disease, I
conclude that there is not one analysis best suited to medicine. The notion of “what
causation is” is context dependent, insofar as the pathologist, the clinician, and the
epidemiologist must (and do) adopt different conceptions of causation.
A complete discussion of the virtues and vices of the conceptual analyses dealt with
in this chapter is a book in itself, so I do not set out to convince the reader to adopt any one
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
59
of the views explicated. I do, however, try to present the most common objections, focusing
in particular on those raised by Kerry et al in the context of EBP; where possible, I respond
accordingly.
Section two looks at a reductive account of causation, according to which causation
is a matter of counterfactual dependence (Mackie 1974; Lewis 1973a, 1973b, 2000;
Hausman 1996). Crudely, the counterfactual thesis takes one event c to be the cause of
another event e if and only if both c and e occur, and if c had not have occurred, then
ewould not have occurred. In this section I argue that the counterfactual account neither
falls foul of the objections raised by Kerry et al, nor of a number of other objections that
may have been raised by philosophers of medicine with similar motivations35. I ultimately
argue that the counterfactual conception of causation is well-suited to the pathologist;
however, although this reductive analysis might be ‘what causation is’ in the context of
pathology, it is not useful for the clinician.
Clinicians must consider the properties of individual patients prior to treatment,
since the qualities of the patient are as relevant to the outcomes of interventions, as the
properties of the interventions themselves. Clinicians often do not know how a patient will
react to a stimulus (be it a potential determinant of disease, or a possible intervention); the
best one can hope for, prior-to-the-fact, is (in general) an understanding of how the patient
is likely to, or is ‘disposed to’ react. In section two, then, following Eriksenet al (2013), and
Kerry et al (2012), I first propose that clinicians employ a dispositional (or ‘tendential’
(Mumford and Anjum 2011)) conception of causation. I then claim that Mumford and
3535
The two additional problems I raise have adapted from common objections in the metaphysics of
causation literature (Mackie 1974; Lewis 1973).
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
60
Anjum’s (2011) vector model of ‘powers’ is, for the most part, a far more accurate
representation of causal reasoning in EBP, than that espoused by the counterfactual
theorists.
To conclude this chapter I outline an unselective regularity account of causation,
according to which causes are non-redundant conditions, together sufficient for their
effects. I demonstrate that Rothman’s (1976) sufficient-cause model to epidemiology is
predicated on this analysis, and that the model is of no use to the clinician. However, I go on
to propose a means of classifying diseases, using the regularity account of causation implicit
in Rothman’s model.
2. Causation as counterfactual dependence
In AnEnquiry Concerning Human Understanding (1748) David Hume defined a cause to be an
“object, followed by another, and where all objects similar to the first are followed by objects
similar to the second. Or in other words, where, if the first object had not been, the second
never existed” (Hume, Enquiry, Sect. VII, Pt. II)36. Although this was not appreciated by Hume
at the time, these are two very different approaches to causation. The first is often referred
to as the ‘naïve regularity account’ (Armstrong 1983; Mumford 2004 ), a more sophisticated
version of which will be presented (in the form of Rothman’s sufficient-case model to
epidemiology) later in this chapter; however, the second is a basic description of the stillpopular counterfactual account of causation.
The counterfactual account states that “If c and e are two actual events such that
ewould not have occurred without c, then c is a cause of e” (Lewis, 1973a, 563). I take this to
36
Oxford University Press (1999) p.146
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
61
be somewhat intuitive. Smoking caused Joe’s lung cancer, since Joe smoked and got lung
cancer, and if Joe had not smoked, Joe would not have got lung cancer. Drinking, on the
other hand, did not cause Joe’s lung cancer, since if Joe had not drunk, hestill would have
got lung cancer – with respect to Joe’s cancer, smoking was a difference-maker (a cause);
drinking was not.
Initial objections to the counterfactual conception
Before outlining the objections to the counterfactual analysis raised in Kerry et al’s defence
of a dispositionalist conception of disease, it is worth presenting twoadditional popular
counterexamples to counterfactual theories of causation.
Objection (i)
Suppose a patient has a major stroke, and dies immediately. Given that she is not, in fact,
simultaneously hit by a bus (or subject to any other life-threatening event), one can
justifiably assume the stroke to be the cause of death. Nevertheless, on the face of it one
can accept that if she had not had a stroke, she mighthave had a heart attack at that same
moment, and died of the heart attack instead. If so, then one implicitly accepts that the
counterfactual “if she had not had a stroke, she would not have died” might be false. The
counterfactual account, then, does not permit one to assert that the stroke certainly caused
the patient’s death (yet it clearly did).
In order to rule out precisely this kind of counterexample, David Lewis goes to great
length to provide a methodology for evaluating counterfactuals. A detailed analysis of his
view is well beyond the scope of this chapter, but crudely: Lewis suggests that there are real
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
62
‘possible’ worlds very similar to ours (the ‘actual world’)37, and that some are more like the
actual world than others. He provides a list of criteria to test how like the actual world (that
is, how ‘close to’ the actual world) these possible worlds are. The list includes a ‘matching of
local particular matters of fact’ clause - again, crudely, if the world appears very similar to
the actual world (in terms of what happens in it), it is likely to be closer to the actual world,
than one that does not resemble it at all.
To evaluate the counterfactual, one considers the closest possible world in which (in
this case) the patient does not have a stroke. This world probably looks exactly like the
actual world until the moment the patient has the stroke. There is then a “minor miracle” in
the possible world, which permits the stroke not to occur. The possible world then runs on
in accordance with the laws of the actual world, and we see whether, in this closest possible
world in which the patient does not have a stroke, she dies of something else instead. If not,
then the counterfactual is true, and one can assert that the patient died of the stroke38.
I take Lewis’s approach to this kind of counterexample to be very effective, and as
we shall see, it resolves a number of problems for the counterfactual conception when
applied to the philosophy of disease. However, talking of other possible worlds, miracles,
and so on, is a little cumbersome in the context of this project. In his 1974, J. L. Mackie
replies to the same kind of objection by introducing an ‘in the circumstances’ clause, to be
expressed prior to the standard counterfactual; that is, he suggests one evaluates whether
37
These worlds are causally isolated from ours, whereby there is a (potentially infinite) plurality of universes.
38
Please note that this short paragraph misses out many of the subtleties of Lewis’s view, and indeed
misrepresents his view in some respects. I aim only to get the very basic gist of his approach to my reader. For
a better understanding of Lewis’s views, refer to his 1973a; 1973b; 1979; 2000. Daniel Nolan’s 2005 provides
an excellent summary.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
63
‘in the circumstances39 [e] would not have occurred if [c] had not’ (Mackie, 1974, 31).
Mackie’s response is, in essence, very similar to Lewis’s. Given the relatively simple way in
which it is expressed, in the remainder of this chapter I speak mainly in terms of evaluating
‘in the circumstances’ counterfactuals. One should note, however, that when I do so, Lewis’s
possible worlds approach is presupposed.
Objection (ii)
The patient’s death during an operation did not cause her to fail to survive, yet given that “If
the patient had not died during the operation, she would have survived it” is true, the
conditional account of causation (as it is expressed here) implies a causal relationship.
Counterexamples of type (ii) occur only because the crude version of the conditional
account outlined here does not assert that causes and effects must be, in Hume’s words,
‘distinct existences’. Mackie suggests ‘it is sufficient to say that someone will not be willing
to say that X caused Y unless he regards X and Y as distinct existences’ (Mackie, 1974, 32).
Similarly, Hausman expresses Lewis’s conception as follows:
L (Lewis’s Theory) c causes e if and only if c and e are distinct events40and if c were not to
occur, then ewould not occur either. (Hausman, 1996, 55)
The patient’s death and the patient’s failure to survive are not distinct existences, since to
not survive just is to die. This second counterexample, like the first, is not overly
problematic for the counterfactual theorist.41
39
My emphasis.
40
My emphasis
41
See also Collins et al (2004)
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
64
Kerry et al’s objections to the conditional analysis in evidence-based practice
In their 2012, Kerry et al defend the view that EBP should assume a dispositionalist
conception of causation – a conclusion that I will later endorse. In the process, however,
they criticise the counterfactual conception of causation, when applied to healthcare more
generally. In this section I critique the objections they raise, as well as some related,
potential objections that they do not. Although prima facie problematic, none of the
objections are convincing (permitting the counterfactual view to be applied elsewhere in
medicine).
Objection (iii)
In the next chapter I discuss the POA, a counterfactual method of effectmeasurementemployed in public health studies. It compares the actual outcome of a study
to some contrary-to-fact supposition. The counterfactual claims the POA permit, however,
must always specify the relevant property/properties of the comparison class; that is,
exposure E causes outcome O versus some contrast class Z (smoking causes cancer relative
to not smoking). In light of this, Kerry et al provide the following objection to counterfactual
accounts of causation in EBP:
In two groups, A (the intervention) and B (the control), there will be a certain proportion
who achieve the outcome of interest, say 58% in group A and 42% in group B.
Depending on the research question, power of the study, etc, statistical analysis will be
performed to determine the significance of this difference. If significant difference is
established, then the recommendation would be that, thus far, A is the preferred
intervention compared with B. In other words, there will be a greater causal effect from
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
65
A than B. But what does this say about the 42% of subjects who responded just as well
with B? Again, the issue is that it looks like something causal did in fact happen to 42%
of subjects in group B, but this cannot be accounted for in the way health science
considers causation. (Kerry et al, 2012, 1009)
The point Kerry et al make here needs clarification, for what they are arguing for is
somewhat unclear42. Suppose I conduct a study, in which I compare a group of smokers who
smoke, on average, 20 cigarettes a day (exposure A), with a group of smokers who smoke,
on average, 10 cigarettes a day (exposure B). The study shows that smoking 20 cigarettes a
day causes cancer versus smoking 10 cigarettes a day (the risk is higher if you smoke more).
If one already knows that smoking 10 cigarettes a day increases risk of cancer relative to not
smoking at all, a study showing that smoking 20 a day causes cancer versus smoking 10
cigarettes a day is interesting, since it provides evidence of what Hill (1965) calls a ‘biological
gradient’43 (the fact that more you smoke, the worse the outcome, provides evidence for
the general causal claim that “smoking causes cancer”), but that particular study tells us
nothing about the causal effect of smoking 10 cigarettes a day versus non-smoking.
Kerry et al are quite right about this. So what is their objection? There are two
possible interpretations. The conclusion of the first interpretation involves making a
category error; the conclusion of the second interpretation, (that the study expressed does
not tell us the causal effect of B simpliciter), is entirely unproblematic.
Interpretation 1: the category error
42
I discuss the methodology Kerry et al are referring to here in far greater depth in chapter 4, section 2
43
See chapter 4, section 1
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
66
Kerry et al, at the end of their critique of the counterfactual approach, write: ‘it is clear…
that the counterfactual conditional fails to get to the essence of what causation is.’
Health studies of the form described above are designed to measure the causal
effects of exposures. They do admit of causal conclusions, but unlike Lewis’s counterfactual
theory, these studies have nothing to do with what causation is. If the objection is that this
‘counterfactual’ approach fails to tell us what causation is, then a category error has been
made.
Interpretation 2: we learn nothing about B
Suppose one wanted to discover the effects of smoking 20 cigarettes a day simpliciter
(strictly speaking, this makes little sense given the structure of epidemiological studies, but
more on that soon) - one would not attempt to discover these effects by comparing a group
who smoke 20 cigarettes a day, with a group who smoke 10 cigarettes a day: one would
compare the 20 a day smokers with non-smokers.
The fact that some people in the study Kerry et al outline responded just as well with
exposure B, as those who responded well with exposure A, implies that both group A and
group B are being treated in some way; but when, in ordinary discourse, one asks “what are
the causal effects of B?” (that is, when one wishes to determine whether or not B is a cause
of an outcome, and how much of an effect it has), strictly speaking (in epidemiological
terms) one is asking “what is the effect of B relative to no intervention”. A well-designed
study of the same format can certainly accommodate this causal question.
In short, of course measuring A relative to B will not tell us anything (or at least, not
much) about the causal effect of B - but that is not the causal question asked by this health
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
67
study. The question asked by the epidemiologist, here, is “how effective is A relative to B?”.
One can find out the causal effect of B relative to no intervention, but that is a different
study.
None of this, however, has any bearing on the success of Lewis’s reductive account
of causation as counterfactual dependence, since the health study’s ‘counterfactual’
approach to effect-measurement is not the same thing,as the counterfactual account of
what causation that they set out to defeat.
Objection (iv)
Kerry et al state that ‘Counterfactual theories… take causes to be the same as necessary
conditions. This would mean however that birth is a cause of death, and having a back is a
cause of lower back pain, that is the counterfactual condition in each of these examples
demonstrates that if you had not been born, you would not have died; if you did not have a
back, you would not have lower back pain’ (2012, p. 1009).
Some might claim that this is, even at face value, unproblematic for the
counterfactual account, since having a back is a cause of back pain.Most, if not all events,
are causally affected by millions of others. Suppose Sam is playing golf, and holes a twenty
foot putt. She strikes the ball sweetly with her putter in the planned direction, at a pace fast
enough to reach the hole, but not so fast that it jumps over – Sam’s putting stroke is a cause
of her holing the putt, but it is not the only causally relevant factor. She might wish to focus
on the good putting stroke, and assert that her stroke was the cause of the ball dropping
into the hole, but the wind, the grain of the grass, the shape of the ball, the slope of the
green, small indentations on the putting surface from footprints and pitch-marks, the force
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
68
of gravity, and many, many other variables feature in determining where the ball finishes.
Were any of these variables different by a significant degree, the ball would not drop. Lewis
states:
We sometimes single out one among all the causes of some event and call it “the”
cause, as if there were no others. Or we single out a few as the “causes,” calling the
rest mere “causal factors” or “causal conditions.” Or we speak of the “decisive” or
“real” or “principal” cause. We may select the abnormal or extraordinary causes, or
those under human control, or those we deem good or bad, or just those we want to
talk about. I have nothing to say about these principles of invidious discrimination. I
am concerned with the prior question of what it is to be one of the causes
(unselectively speaking). My analysis is meant to capture a broad and
nondiscriminatory concept of causation. (Lewis, 1973, p. 558–9)
For those, like Lewis, who refuse selective theories of causation, (iv) looks to be
unproblematic. But medicine is surely not at liberty to adopt such a conception - not many
doctors would consider the presence of the aorta to be a cause of heart disease!
Fortunately, there are selective accounts of causation on offer (Menzies 2004;
Schaffer 2005; Broadbent 2008). Again, a long discussion of causal selection is beyond the
scope of this chapter - but let us briefly consider two options: Mackie explains that:
…causal statements are commonly made in some context, against a background
which includes the assumption of some causal field. A causal statement will be the
answer to a causal question, and the question “What caused this explosion?” can be
expanded into “What made the difference between those times, or those cases,
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
69
within a certain range, in which no such explosion occurred, and this case in which
an explosion did occur?” Both cause and effect are seen as differences within a field;
anything that is part of the assumed (but commonly un-stated) description of the
field will, then, be automatically ruled out as a candidate for the role of cause.
(Mackie, 1974, p. 34-5)
Mackie thus agrees with Lewis insofar as he takes causes to be difference-makers, but
disagrees insofar as he does not adopt his ‘broad and non-discriminatory’ approach. Causal
questions presuppose a causal field44; that is, a set of circumstances, no part of which can
be deemed a cause. When one asks ‘what caused Bob’s lung cancer?’ the existence of his
lungs, his being alive, and so on, all form a part of the field, and soex hypothesi cannot be
causes. Bob’s smoking, on the other hand, does not form part of the causal field, so if the
conditional states that if he had not smoked, he would not have had lung cancer, one can
identify smoking as a cause of his cancer.
Jonathan Schaffer (2012) defends ‘causal contextualism’ (versions of which are also
defended by Woodward 2003; Maslem 2004), the thesis that: ‘A single causal claim can bear
different truth values relative to different contexts, where this difference is traceable to the
occurrence of ‘causes’, and concerns a distinctively causal factor’ (Schaffer, 2012, 37). One
might ask, for example:
1. Why did these elephants cross the river?
The emphasis on ‘these’ implies that one is questioning the cause of this specific herd of
elephants crossing the river, rather than the herd next to them. Perhaps the answer is:
44
How the question distinguishes between the conditions and the causes is unclear.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
70
“because there wasn’t enough food for the two herds, and this was the only herd able to
cross the river”. On the other hand, one might ask:
2. Why did this these elephants cross the river?
Here, the emphasis suggests one is questioning why the herd of elephants crossed the river
rather than, say, staying where they were. Perhaps the answer is “because there were no
trees further upstream”, but it is not “because there was not enough food for the two herds
of elephants…”
For Schaffer, causal claims are sensitive to contrasts, whereby “what shifts with
context are the contrasts in play, where contrasts are specific possible alternatives to actual
events”. (p. 45). With scenario 1, the contrast classes involve the other herd of elephants
attempting to cross the river; with scenario 2, the contrast classes involve the herd of
elephants that crossed the river, walking upstream, staying where they are, moving inland,
and so on.
The contrastive view, he writes, can “be plugged into a simple counterfactual
[conditional] test for causation by replacing the supposition of the non-occurrence of c or e
[cause or effect], with the supposition of the occurrence of the associated contrast [c* or
e*]. So, for instance… one might hold that c rather than c* causes e rather than e* iff
(roughly) if c* had occurred then e* would have occurred.” (p. 45-46).
Kerry et al claim that the counterfactual theory implies that having a back is a cause
of back pain - but if Schaffer’s view is coherent, the counterexample can be defeated, since
no contrast class for any reading of the question “why does this man have back pain”
involves a man with no back. I do not pretend to have provided a detailed critique of either
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
71
Mackie’s or Schaffer’s conception. I outline the views here only to show that theories of
causal selection exist, and that they are not prima facie implausible.
Objection (iii)
Suppose that, in the same patient, at the same time, two events (say, A and B) occur that
are individually sufficient for the same disease D. Neither is a necessary condition, since
neither the counterfactual “if A had not occurred, D would not have occurred”, nor the
counterfactual “if B had not occurred, D would not have occurred” is true. Therefore, given
that causes are necessary conditions, neither A nor B is a cause of D.
This is a common occurrence - effects are often causally overdetermined. Supposea
firing squad of four shoot a soldier. It is not true of any member of the firing squad that, if
he had not pulled the trigger, the soldier would still be alive. It would clearly be very
damaging to the counterfactual account, if it implied that none of the bullets killed our
soldier – he is dead, after all. One can, however, deny that any one of the bullets individually
caused the soldier’s death, and suppose that all the bullets, together, caused the soldier’s
death (for it is true that if none of the bullets hit the soldier, then the soldier would not have
died).
Rather than considering either A or B causes individually, one must take the cause of
the disease to be A and B together. The counterfactual “if neither A nor B occurred, then D
would not have occurred” is true, so the problem disappears.
Objection (iv)
“If [a randomised controlled trial (RCT)] failed to show a significant difference between two
intervention groups, but in both groups a treatment effect was observed, then the
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
72
counterfactual stance would have to support the statement that neither intervention
caused the effect.” (2012, p. 1009). This issue, I believe, is either illusory, or can be resolved
with Mackie’s ‘in the circumstances’ clause.
First, a brief explanation of a basic RCT: a standard RCT involves two maximally
causally homogenous experimental groups (crudely, the groups are causally homogenous
insofar as they would produce the same outcomes, from the same exposures, for the same
reasons, were neither group subjected to the exposure the RCT is investigating), from a
subpopulation of the target population (the population to which the results are to be
‘exported’ - perhaps, but not necessarily, the general population (Cartwright 2007)). One of
the experimental groups is then exposed to some humanly manipulable intervention (e.g. a
drug, a particular diet, etc.). After a specified period of time, the two groups are compared,
and given that they were otherwise causally homogenous, any significant differences are
the effect of the exposure. In terms of counterfactuals: suppose there are two experimental
groups in an RCT, group A, and group B. Group B is exposed to Eand group A is not. If the
risk (statistical frequency)45, R1, of some outcome O is higher in group B at the end of the
study, than the risk, R2, of O in group A, then one establishes a causal connection between E
and O, since if group B had not been exposed to E, then R1 would be equal to R2. E was a
difference-maker, and hence a cause.
Unlike the simple case above, some RCTs involve two or more intervention groups.
In a case where there is no control group, one can only compare the results of the two
intervention groups. Of course, without reference to any other data, this tells you very little.
45
Risk is a statistical notion. The risk of an outcome is the number of people in the experimental group in which
the outcome obtained/the number of people in the experimental group at the start of the study.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
73
After all, when the outcomes are the same, the two intervention groups either show the
same causal effect, or no effect whatsoever. In this type of trial, just as with objection (iii),
the question asked by those conducting the studies is merely a comparative one concerning
effect-measurement, and hence to use this as a criticism of the counterfactual analysis of
causation is entirely misguided.
Sometimes there might be two interventions and a control group. In these cases,
there are in effect two distinct RCTs (let us call these RCT1 and RCT2) being conducted. Each
RCT has an exposure group (X1 and X2 respectively), and they share an otherwise causally
homogenous control group (Y). RCT1 tests exposure A in X1, and RCT2 tests exposure B in
X2. A and B are distinct, yet produce the same causal effects in X1 and X2 respectively (that
is, the differences in outcome between X1 and Y, and between X2 and Y, respectively, are
identical). One might claim that because the causal effects of A and B are identical, neither
can be deemed a cause, but again, this would be misguided. If the trial has been well
designed, then when one adds the ‘in the circumstances’ clause (which fixes the control
group Y) the right counterfactuals come out true: “in the circumstances, if (contrary to fact)
group X1 had not been exposed to A, then the outcomes of X1 and Y would not have
differed’ – A is thus deemed a cause according to the conditional account, since it was a
difference-maker.
Conclusions
Kerry et al’s conclusion is that it “is clear… that the counterfactual conditional fails to get
across the essence of what causation is46…”(p. 1009). It is far from clear, though, that any of
46
My emphasis
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
74
the arguments presented in this chapter support it. Lewis’s conception of causation as
counterfactual dependence is, at least on the face of it, a fairly robust reductive account of
‘what causation is’. Furthermore, the view is very well suited to the pathologist, whose job is
very often to find and monitor the cause of a patient’s illness, or in some instances, the
cause of death (that is, to find C in the counterfactual: Were the patient not to have had
condition C, she would not have died). Nonetheless, the counterfactual approach is not well
suited to all medical contexts.
In chapter 4 I outline in more detail how the epidemiologist discovers the causal
effects of interventions, focusing on the POA. We shall see that the methodology of this
theoretical framework, like RCTs, draws heavily on the counterfactual account of causation
– but the experimental justification of general causal claims, although (of course)
fundamentally important to EBP, does not reflect the day to day activities and thoughtprocesses of the clinician. Clinical medicine focuses on the signs and symptoms of individual
patients, and how best to treat the inferred conditions. Answering the clinician’s questions
requires a new conceptual framework. In the next section I outline a dispositional
conception of causation, and show how modelling causation in terms of dispositions is more
useful to EBP than the counterfactual conditional approach.
2.1 Clinical medicine, and the dispositional account of causation
Just as salt is disposed to dissolve in water, and fragile objects are disposed to shatter,
humans are disposed to contract diseases. Many of our physiological subsystems are
disposed to dysfunction, and these dispositions can be triggered by the dispositions of
invading disease entities (human papilloma virus is disposed to cause genital warts, for
example), by those of our internal organs, by those of chemical substances (poisons), and so
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
75
on. Epileptics are disposed to have seizures after excessive alcohol consumption; that is,
alcohol is a trigger, or ‘stimulus condition’ for seizures, which ‘manifests’ when consumed
by epileptics (in the absence of interfering factors). Some are disposed to go into
anaphylactic shock when stung by bees, and smoking can trigger one’s disposition to get
lung-cancer.
The dispositional conception is a nonreductive account of causation (Mumford and
Anjum 2011) that in this context, takes individuals to ‘tend towards’ disease, and the
determinants of disease to ‘tend towards’ causing disease (Eriksenet al 2009; Kerry et al
2012). In all the above cases, the stimulus conditions – viruses, poisons, alcohol, bee stings,
smoking, sugar-ingestion, and so on – interact with the dispositions of patients, and can, but
do not necessarily, lead tochanges in the functional efficiency of physiological subsystems.
In short, (i) every physiological change that takes place whilst diseased (indeed, any
physiological process at all!) seems to beexplicable using dispositional terminology; and (ii)
the justification for clinical decisions seems to be, at least partly, grounded by what the
clinicians see to be potential risks and benefits of the prescribed interventions; which, again,
are explicable in terms of dispositions.
Unlike the counterfactual account, this view seems well-suited to clinical practice.
Humans are disposed to contract malaria when bitten by certain malaria carrying
mosquitos. Clearly, in the circumstances considered ‘normal’ in non-malarial zones, this is
not a problem, since no malaria carrying mosquitos are around to trigger that disposition but of course that can change when one enters Kruger National Park. Jane (like most other
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
76
people) is disposed to contract malaria if bitten by a malaria carrying mosquito47, and given
the presence of these mosquitos in Kruger National Park (depending on the time of year and
where one is in the park), she might contract the disease when she visits. The doctor cannot
remove the disposition of mosquitos to spread malaria, but by prescribing the prophylactics,
she can remove Jane’s disposition to contract malaria prior to her entering the park.
The Mumford-Anjum model
In some regions, an important aspect of clinical medicine is the treatment of poisons using
antidotes; again, this can easily be accommodated by a dispositional conception of
causation.Suppose Rachel is bitten by a snake and taken to hospital for treatment. Using the
vector model of ‘powers’ proposed by Mumford and Anjum (2011, 39)48, figure 5 represents
the reaction Rachel’s physiological response to snake venom after an antidote has been
injected. The vectors represent causes/conditions that dispose Rachel either towards
health, or towards disease. The length of the arrow determines strength of that
contribution, and the direction of the arrow determines whether the condition disposes
towards health, or towards disease. Take a to represent the disposition of the brain to bleed
when subjected to snake venom, b to represent the disposition of her intestines to be
damaged by snake venom, and c to represent the dispositions of the snake venom when
injected into the blood. These causal factors (dispositions) all contribute towards the
47
It is perhaps worth noting (purely out of interest) that relatively few species of mosquito can transmit
parasites of the genus Plasmodium (that which causes malaria in humans) – the most common genus of
malaria carrying mosquito, Anopheles,includes over 450 different species, only 30-40 of which can transmit
malaria to humans.
48
Mumford and Anjum do not think dispositions and powers are interchangeable concepts, since according to
them, dispositions are ‘clusters of powers’ (Mumford, 2004, 168-179; Mumford and Anjum, 2011, 99-103).
However, in the philosophy of medicine literature, philosophers focus on dispositionality (Scheuermannet al
2009; Kerry et al 2012), and Mumford and Anjum’s vector model is equally useful for modelling dispositions.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
77
dysfunction of Rachel’s physiological subsystems, that is, towards disease. Now take d to
represent the pre-existing dispositions of Rachel’s physiological subsystems to ‘tend
towards’ health, and e to be the dispositions of the antivenin. R represents the resultant
vector, which indicates whether, given all the causally relevant states of affairs, Rachel is
disposed towards health, or towards disease:
Figure 5:
The Mumford-Anjum model
Given the injection of the antivenin, R shows that Rachel is disposed to recover, since the
sum of the vectors that contribute towards health, outweigh the sum of the vectors that
contribute towards disease. If she had not been injected with the antivenin e, however, the
sum of a, b, and c (that is, the dispositions of her body to be affected by snake venom, and
the dispositions of the snake venom itself) outweigh d (the dispositions of Rachel’s
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
78
subsystems, under normal circumstances, to maintain a healthy state), and so Rachel would
have been disposed towards disease. The strength of the disposition towards disease or
health is thus represented by the size of the resultant vector.
If one has a ‘chronic disease course’ or a ‘progressive disease course’49
(Scheuermannet al), then, the task of the clinician is to make the resultant vector (which, if
it has magnitude, must point towards disease) as small as possible; and, if the disease can
be cured, the task of the clinician is to make sure the resultant vector points towards health,
with as great a magnitude as possible - to tend towards a quick recovery. Note that the
model can represent any two properties. One might, for example, be interested in reducing
pain, in which case, the vectors modelled would point not towards health and disease, but
towards pain and comfort. Of course, just because R tends towards comfort, or towards
health, that does not guarantee comfort or health – but this is a virtue of the thesis.
Adispositionalist reading of cause allows one to assert the (intuitively correct) general causal
claim that ‘arsenic causes death’, since the dispositional quality of that claim (that is, the
fact that death is not necessitated by arsenic) is implicit in the assertion.
The Mumford-Anjum model is concerned with the dispositions of individual patients,
and the effects of interventions on those individuals (as opposed to the effects of
interventions on populations); the model can represent whatever qualities the clinician and
patient are interested in; and the model accommodates the fact that clinicians regularly do
not know how their patient will react to interventions during a disease course – they must
49
Whereby either the ‘totality of all processes through which a given disease instance is realised’… (a) does not
terminate in a return to normal [health] and (b) would, absent intervention, fall within [a disease] range… [or]
(a) does not terminate in a return to normal [health] and (b) would, absent intervention, involve an increasing
deviation from [health]. (Scheuermannet al, 2009, 118),
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
79
operate on the basis that the right interventions dispose the patient towards health. The
vector model, then, looks to be an excellent representation of EBP.
3.The classification of diseases, and the sufficient-cause model of causation
In this section I discuss the sufficient-cause model of causation, and the related regularity
model proposed by Mill and developed by Mackie50. This model does not represent the
practices of the clinician, since it is not a tendential account. Nevertheless, that the
sufficient-cause model does not reflect clinical practice does not make it misrepresentative
in the stronger, metaphysical sense; that is, the tendential concept of causation employed
by clinicians is not deterministic, but that does not imply that as a matter of
fact,determinism (in medicine, and more generally) is false. It may be the case that clinicians
must employ a tendential model, but nonetheless true that certain sets of conditions do
necessitate disease states. Given that medicine is a pragmatic discipline, one might think
this metaphysical claim somewhat irrelevant. However, outlining this view is not a fruitless
task, since the sufficient-cause model is central to the method of disease-classification I
propose in section 4 of this chapter.
Consider Jane, a Kruger National Park safari guide - Jane is a little laissez faire about
prophylactics, but when she feels there is a high risk of malaria, she takes them nonetheless.
When deciding whether to take her pills, she considers the following factors:
(i) When she is working. In mid-winter, there are very few malaria carrying mosquitos
anywhere in Kruger National Park.
50
And later developed by Mackie.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
80
(ii) Where she will be based - when there are mosquitos in the park, they are restricted
to the northern region.
(iii) The quality of the insect repellent - even if there are lotsof malaria carrying
mosquitos around, she will not be bitten, given an effective insect repellent.
With this in mind, she creates the following model:
Figure 6:
Malaria’s ‘sufficient-cause’
Jane believes (quite reasonably) that diseases are not usually determined by a single
cause. She also believes that if criteria (i) to (iii) obtain, and she does not take prophylactics,
she will get malaria – the conditions represented by figure 6, then, are together sufficient
for contracting the disease. Furthermore, Jane’s model represents what she deems to be
the only jointly sufficient conditions for malaria (which in this case, rightly or wrongly, are
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
81
translatable to what she deems to be the jointly sufficient conditions for being bitten by a
mosquito).
If conditions (i) to (iii) are met, then Jane will take prophylactics, since if she does not
the sufficient-cause of malaria obtains – in taking the prophylactics, she removes a segment
of the pie-chart, and so the sufficient-cause does not obtain, and she does not contract
malaria.Jane is using Rothman’s ‘sufficient-cause model to epidemiology’ (Rothman 1976;
Rothman et al, 2008, 8) (of course, there may be many pie-charts for any one disease - there
is more than one way to catch hypothermia - but every disease has at least one such chart).
As I have already stated, any suggestion that Rothman’s model accurately represents
clinical practice would be misguided. There are two main reasons for this.
(i) Either one deems all the conditions mapped in the pie-charts to be causes, or one
takes the cause of disease to be the entire sufficient-cause. Neither, of course, is
compatible with clinical practice.
Rothman’s model is unselective, so (once again) the ‘presence’ of lungs becomes a cause of
lung disease. In these cases, however, one cannot appeal to either the causal field or
contrast class responses. Causal fields and contrast classes are fixed by the nature of the
causal question, and Rothman’s sufficient-cause model does not ask causal questions; it
merely provides information concerning jointly sufficient conditions for contracting token
diseases.
(ii) The model is inherently deterministic.
As detailed in the previous section, clinicians cannot operate in this way. Clinical medicine,
at least on the face of it, presupposes a dispositional conception of causation.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
82
Rothman’s model, despite not resembling the clinician’s concept of cause, is
nonetheless important for both clinical medicine and epidemiology. In the next chapter I
discuss theoretical approaches to causal inference and effect measurement in
epidemiology, concluding that Rothman’s model is essential if one is to capture
nonmanipulable causes in one’s epidemiological framework. Here I show how the model
(although expressed in a slightly different way) can ground an effective means of classifying
diseases.
4.1 On the classification of diseases51
Identifying the necessary and sufficient conditions for individuating and classifying diseases
is a matter of great importance in the fields of law, ethics, epidemiology, and of course
medicine. Here I propose a ‘causal classification of disease’ (CCD). Note that this is not a
metaphysicalaccount of what a disease is, either from an ontological or conceptual
perspective, but an account of how to individuate one disease from another (the cause of a
disease is not the disease itself!).
When a patient becomes ill with influenzaF one might determine the contraction of
the influenza virus V to be the cause. However the virus is not the only factor involved in the
patient contracting the disease. She has had a particularly poor immune system since she
developed AIDS,P. This contributed to her contracting F, as ceteris paribus without P her
immune system would have been strong enough for her to fight off the infection. Neither V
nor P is individually sufficient for F, but both are causally relevant and jointly sufficient.
51
With kind permission from Springer Science+Business Media: Theoretical Medicine and Bioethics, On the
Classification of Diseases,2014, August;35(4) pp.251-69
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
83
Despite their being individually insufficient, we might nonetheless consider the cause to be
F, or P, or both, depending on what information we consider most salient.
Furthermore, what appears to be the ‘salient’ cause often acts as an etiological
agent for more than one disease (sometimes, in the same patient, at different times).
Consider a case where the HPV virus causes a patient to manifest genital warts as a young
woman, and at a later date the virus is deemed to be the cause of her cervical cancer.
Although the two distinct diseases are caused by the same etiological agent, one can see
that the invading organism does not tell the whole causal story.
Figure 7:
Multifactorial accounts of disease
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
84
According to ‘multifactorial accounts’ (Broadbent 2009), every token disease has
numerous causes, that are only jointly sufficient for it52. Furthermore, there are likely to be
numerous distinct jointly sufficient conjunctions of causalconditions. This account is thus
captured by Rothman’s sufficient-cause model. That model itself draws heavily on Mackie’s
1973 ‘regularity account of causation’, which I use here to provide a means of individuating
diseases.
When we consider what the causes of cardiovascular disease are (on a more general
scale), we have a disjunction of conjunctions of causalconditions. Cardio vascular disease
(CVD), for example, can result from ‘XY or QY or….’ Where ‘…’ represents a finite number of
conjunctive conditions; we may assert that CVD is always preceded by ‘XY or QY or…’.
Conversely that all ‘XY or QY or…’ are followed by CVD. It is thus a regularity account in
Hume’s sense, insofar as a cause is constantly conjoined with its effect.
A disease obtaining might also be partially reliant on the absence of certain
occurrences. In the CVD case, for example, lots of exercise might ‘mask’53 the disease. These
are welcome additions to the causal classification model, as many diseases, such as scurvy
and rickets, are the result of a lack of something (in these cases, vitamin C and vitamin D
respectively). The result picture looks like this: ‘Z iffXYnot-D or QYnot-D or...’54.This model is
52
For now, consider a causal condition to be an event causally relevant to the contraction of the disease in
question.
53
‘Mask’ is used here in Johnson’s (1992) sense – exercise might mask a patient’s disposition to get CVD just as
dampening a match masks its disposition to light when struck.
54
This is integral to J.L. Mackie’s (1974) conception of causation, where Mackie identifies a cause as any
“insufficient but non-redundant part of an unnecessary but sufficient condition” (an inuscondition), that is, any
one condition to be found within what he terms, the ‘full cause’.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
85
clearly unselective, but in the context of disease-classification, this is a virtue, not a
problem.
Given the removal of ‘selection’ from the process of cause-identification, the
exhaustive list of conditions to be satisfied for most effects (for our purposes these are
diseases, but this is a general account of causation) is enormous. Mill writes, though, that
contrary to our every-day conception ‘the cause…philosophically speaking, is the sum total
of the conditions positive and negative’55. Following Mackie, let us call this ‘philosophically
speaking’ cause, the ‘full cause’.
According to CCD, diseases are to be individuated by their full cause; that is, the
disjunction of conjunctions of events jointly sufficient for contracting the disease.
Pragmatically speaking, the ‘full cause’ is clearly not what we refer to in standard
causal talk, either within or outside of an epidemiological/medical context. When we
identify causes in a medical context the salient information is that which leads to diagnosis,
prognosis, explanation, and the identification of suitable treatments for the disease. Indeed,
a medic need not even identify precisely what disease a patient is suffering from in order to
(non-accidentally) prescribe the correct treatment. Many antibiotics, for example, will
successfully treat a large number of different bacterial infections, so in many cases to help a
patient the medic need only identify the infection as bacterial. The salient cause(s) (where
cause is being used in the every-day sense of the term; that is, one of the conditions in one
of the conjuncts that partly comprise our full-cause) and corresponding treatment(s) are
thus often associated with many different diseases.
55
System of Logic, Book III, Ch. 5, Sect. 3.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
86
Given that one etiological agent can apply to many different diseases, using a single,
salient causal factor (inuscondition – see footnote 54) is not sufficient for a causal
classification of disease. That said, once we take the cause to be the cause qua ‘cause
philosophically speaking’, no two diseases will have the same full cause.
4.2 Objections to the causal classification model
Objection(i)
Mackie’s general account of causation implicitly assumes that our conceptual framework is
deterministic, but it’s not obvious that a conceptual account of disease individuation can be
so. Epidemiology is rife with talk of chance: “chance of infection”, “chance of survival”,
“chance of going into remission”, and so on.
Response:
Chance, in medical contexts, tends to refer to risk, whereby one’s assigned
chance of survival/recovery is determined by the percentage survival rate of similar patients
with similar ailments, at similar stages of the disease, in similar environmental conditions,
and so on. Frequentist probabilities do not rule out deterministic conceptions of causation
in medicine - ‘smoking raises the (statistical) probability of getting lung cancer’ and ‘lung
cancer has its identity fixed by a (deterministic) full cause’ are perfectly compatible claims56.
Objection (ii)
Different stages of a disease have different ‘full causes’, so CCDwould identify them as
different diseases.
56
There are also possible ways of amending the account to allow for real chance (that is, allowing that even
upon knowing all states of affairs at t and all the laws of nature, it would be impossible to know for certain the
state of affairs at t+1 due to objective probabilities in nature). One can give probability-raising accounts of
causation, where the causal conditions are such that they are necessary to raise the probability of the effect in
the circumstances. This would, in fact, be sufficient for my thesis. However, here I am happy to endorse
metaphysical determinism and frequentist accounts of probability.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
87
Response:
The model identifies diseases by the causal conditions of their initial
contraction. Any stage of a disease could be traced back to the conjunction of events that
caused the first stage of the disease – CCD individuates a disease by the full cause only of its
first stage (that’s not to say that a later stage of a disease cannot be individuated by the full
cause of a disease-stage, but this is a virtue of the model).
Objection (iii)
Assuming determinism, and the transitivity of causation, the cause of any disease in person
P is also the cause of any subsequent diseaseP might contract. Now supposeP has infectious
disease D, which weakens her immune system, and as a result she contractsdiseaseD*. For
practical purposes it is (let us suppose) useful to treat D and D* as distinct diseases, but how
can CCD accommodate this, when it implies thatD and D* can be seen as different stages of
the same disease?
Response:
The aim here is to find a means of individuating diseases, that is, for no two
diseases to satisfy the same conditions - CCDsucceeds in this regard. CCDdoes not tell us
which full causes identify diseases and which do not, but if D and D* are distinct diseases
(which we have agreed they are), then they each have their own unique full cause. Although
it raises an important issue, objection iii does not defeat CCD.
Objection (iv)
“…if all we say about diseases is that they have many and diverse causes, it is not clear what
more we can hope for than a catalogue. And the catalogue will never be complete, because
in practice the causes of even the most ordinary event are beyond human counting”
(Broadbent, 2009, 308). Here Broadbent makes the point that the full cause (or something
like it) cannot be used in general explanation, and that since diseases are ‘kinds of ill health
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
88
for which general explanations are available’, a catalogue of causes is an inadequate means
of classifying disease.
Response:
I agree with the motivations for Broadbent’s objection. We classify diseases
for practical purposes, primarily in the hope that we can better cure and/or prevent illhealth. To do this we first need the model to provide general explanations for ill-health, but
does CCD fail in this regard? Broadbent claims that a model of disease purely comprised of
lists of causes does not explain why we differentiate between different diseases, ‘it says
nothing about what disease is, nor about why we distinguish between diseases in the way
that we do57’. However in token cases (and here I speak of causal sequences in general, not
just those concerning disease), causal explanations of particular events can be very useful.
With token cases of ill-health, causal explanations can undeniably help us determine which
procedures should be followed to restore good health58. So why not generalise this claim,
that is, from the cause of a disease-token to the full cause of a disease-type? If we identify a
disease by its causes, we explain the ill-health – diagnosis becomes the identification of
causal explanation (more specifically, the identification of the range of possible causal
explanations), and points towards possible treatment(s) for the disease-type diagnosed.
Furthermore, Broadbent’s objection is raised against bare multifactorial accounts – I do not
propose that all there is to a disease is a list of causes; rather, I take diseases to be the
failure of physiological subsystems to perform their natural functions, and these diseases
can be individuated by lists of causes.
With respect to the second of Broadbent’s worries (that ‘the causes of even the
most ordinary event are beyond human counting’), it is indeed pragmatically impossible to
57
From personal correspondence with Alexander Broadbent, 2013
58
More on this in section 4.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
89
know all the causalconditions and their ordering for any one disease, but the practical
inability to fully pick out all the identity conditions for a disease does not significantly
jeopardise the position. The worry is that diseases become ‘unknowable noumena’, but this
is not an uncommon phenomenon in the sciences.Brian Ellis (2001, 2009)59 takes the aim of
science to be to discover the essential properties of natural kinds, but not knowing all the
essential properties of a particular natural kind does not entail that the natural kinds are
unknowable, nor does it make scientific essentialism wholly unappealing. Just as discovering
unknown essential properties of natural kinds refines our scientific knowledge and
understanding of the natural world, discovering further possible causes of a disease
improves our knowledge, understanding, and ability to manipulate the disease. We may
have to concede that knowing an individual disease’s full cause/identity conditions is
beyond the scope of human understanding, but this does not make the position implausible.
5 Conclusions
In this chapter I have considered three conceptions of causation: causation as
counterfactual dependence; dispositionalism; and a sophisticated form of the regularity
conception of causation. Each seem to be appropriate for some areas of medicine, but not
for others.
When one questions whether one event caused another, Lewis’s counterfactual
conception seems to capture the issues at stake. Did the proposed cause actually happen,
and would the effect occurred had that proposed cause not? This conception, it seems to
me, is very likely to be the that employed by pathologists when deciding cause of death, and
perhaps in the diagnostic stages of clinical practice. Furthermore, the objections to the
59
See also Ellis and Lierse (1994)
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
90
counterfactual theory raised in this chapter are all ultimately unconvincing, so there is little
reason to remain sceptical of the view in the contexts identified. However, the
counterfactual approach is ill-equipped to deal with clinical practice.
The clinician is in the business of explanation and prediction; she must explain why a
patient is feeling ill, and having explained the illness, predict which interventions will return
her patient to a state of good health, in the fastest possible time. The counterfactual
account of causation may be sufficient for the explanatory aspect of the clinician’s role,
since this is to provide ‘after-the-fact’ causal explanations. However the predictive side
cannot be dealt with in this way.From a conceptual perspective at, least, interventions do
not seem to necessitate outcomes. Furthermore, patients with the same disease will very
often react differently to the same medication, since the outcome of an intervention
depends as much on the properties of the patient as it does on the properties of the
medication (penicillin will cure Jim’s chest infection, but kill John, for John is allergic to
penicillin). In clinical medicine, it seems, a nonreductive, dispositionalist conception of
causation should be adopted.
The sufficient-cause model is useless to the clinician; however, both Rothman’s
conception, and the Mill/Mackie account of causation from which it is derived, are very
important to medicine. In the next chapter I show that, unlike the POA, the sufficient-cause
model successfully accommodates nonmanipulable causes in epidemiology. In this chapter I
showed that the full/sufficient cause of a disease, might well ground a useful method of
disease-classification.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
91
Chapter 4. Population Health and Causal Inference
1.1 Identifying causal relationships: causal questions for epidemiologists
In the introduction I defined epidemiology as the study of disease in populations, for the
purposes of improving public health. This project is inextricably tied to the concept of
causation, since in order to improve public health, one must (amongst other things, of
course) identify the determinants of disease. To this end, epidemiologists have developed
many types of experimentaland observational studies, and a wide variety of statistical
techniques. Correlations are relatively easy to discover, but discovering a correlation is not
the same discovering a causal relationship. For example, drinking is correlated with lung
cancer, but it does not cause lung cancer. It is correlated with lung cancer because drinkers
are more likely to smoke, and smoking causes lung cancer. So a major part of the
epidemiologist’s job, is to determine which correlations are causal, and which are not.
Furthermore, epidemiologists do not merely wish to establish causal relationships between
exposures and outcomes, but how strong those causal relationships are; that is, how much
of a difference the exposures makes to the outcomes60.
This chapter is concerned with the theoretical nature of causal inferences and effect
measurement in epidemiology. Following Bird, in the next section I outline Hill’s nine criteria
60
One of the epidemiologist’s roles is to advise how best to allocate funds - in principle, one allocates the most
money to dealing with exposures with very strong causal connections to common and devastating diseases.
For example, epidemiologists discovered a very strong causal connection between smoking and lung cancer,
and a relatively very weak (but nonetheless genuine) causal connection between air pollution and lung cancer.
Unsurprisingly, more money is spent trying to get people to stop smoking, than trying to get people to avoid
walking along busy roads.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
92
for causal inference in medicine, demonstrating some of the problems epidemiologists face
in making causal inferences. I then present two models of causal inference commonly
employed by epidemiologists. First, the POA is put forward as a means of identifying
practicable (Woodward 2004) causes of disease, and measuring the strength of the effects
of exposures; I then question how successful this approach to causal inference in
epidemiology really is. We shall see that the POA does not accommodate all causal factors
relevant to epidemiological practices; to resolve this issue, I suggest supplementing the POA
with Rothman’s (1976) sufficient-cause model.
1.2 Hills criteria and the causal inference problem
Epidemiologists use a number of different kinds of study to identify causal relationships. The
majority of these studies are ‘observational’ studies, such as cohort61 or case-control
studies62, but merely observing correlations in not enough to infer causation.Here I outline
Hill’s famous 1965 list of nine criteria for distinguishing between causation and mere
correlation or association (Hill, 1965, pp. 295-300); however, we shall see that these criteria
are ofteninsufficient to establish causation.
1. Strength.
Hill provides two examples to illustrate that the stronger the
association between exposure and outcome, the more justified one is in identifying a
causal relationship: ‘the mortality of chimney sweeps is 200 times that of workers
61
In a cohort study, two groups are chosen at the beginning of the study, one with the exposure (e.g. smoking,
a kind of medication, etc) and the other without. After a specified period of time, the epidemiologist
investigates the proportion of those groups with the outcome (e.g. cancer, health, a side effect, etc), and the
results are compared.
62
In case-control studies, a group with ‘the outcome’ (e.g. cancer) is studied, to determine in what proportion
of the group the exposure obtained (‘the risk of exposure’ – e.g. what proportion of the cancer patients
smoked). This result is compared with the ‘risk of exposure’ of a group of people in which the outcome does
not obtain(e.g. a group of people not suffering from cancer).
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
93
who were not specially exposed to tar or mineral oils’; and ‘the death rate from
cancer of the lung in cigarette smokers is nine to ten times the rate in non-smokers’.
2. Consistency. This important criterion emphasises the need for causal associations
to be repeatable, not only in the same location and at similar times, but under a
wide variety of circumstances.
3. Specificity.
The more specific/well-defined the exposures and outcomes, the
more justified one is in inferring a causal relationship.
4. Temporality. ‘Which is the cart and which is the horse?’ In order to distinguish
between cause and effect, one must consider which event precedes the other.
5. Biological gradient.
If there is a causal relationship between the exposure and
outcome, often there will be an observable biological gradient. In the case of
cigarette smoking, for example, the more cigarettes one smokes, the more likely one
is to get lung cancer.
6. Plausibility.
The biological plausibility of a causal inference is determined by the
accepted biological theories of the time. Of course, what is believed by scientists
often changes over time, so consistency with the current popular theories in biology
is not a necessary condition for inferring a causal relationship, but it is nonetheless
desirable.
7. Coherence.
Although biological plausibility cannot be demanded, an inferred
causal relation ‘should not conflict with the generally known facts of the natural
history and biology of the disease’.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
94
8. Experiment.
Hill writes: ‘Occasionally it is possible to appeal to experimental, or
semi-experimental, evidence. For example, because of an observed association some
preventative action is taken.’ He asserts that this provides the strongest support for
causal inference, and indeed, controlled trials like RCTs are typically deemed the
most reliable means of establishing causation.
9. Analogy.
If two exposures, x0 and x1 are sufficiently similar, and there is a
known causal relationship between x0 and outcome O, it is sometimes reasonable to
infer a causal relationship between x1 and O.
Hill does not claim that any of these criteria (with the possible exception of specific kinds of
well-designed and well-conducted experiment) individually succeed in identifying causeeffect relationships, but that each provides partial support for a causal link obtaining is fairly
intuitive. Nonetheless, particularly in the case of observational studies, even close
adherence to Hill’s criteria will often result in spurious conclusions.
In an observational study, the epidemiologist records information about the subjects
under study, taking measurements and asking questions, without actively intervening in
their lives, or in their environments (Saracci, 2010, 10; Broadbent 2013, 5-7). Here I show
how Hill’s criteria are applied to observational studies, but conclude that, even when these
criteria are fulfilled, the observational studies so often employed by epidemiologists, are not
always conclusive.
In order to establish a causal theory (T) – that there is a causal relationship between
an exposure E and outcome O - one must eliminate: (R) the reverse hypothesis (that O
causes E); (C) all competing hypothesis (e.g. common causes); and (N), the null hypothesis
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
95
(Bird 2011). Bird argues that, in observational studies, Hill’s criteria of strength, consistency,
specificity, and biological gradient, support (T) by eliminating the null hypothesis. In other
words - strong correlations between E and O; correlations between E and O in different
circumstances; well-defined exposures and outcomes; and observable biological gradients –
all increase the likelihood that the NHST will deem the correlation significant.
The temporality condition in an observational study supports T by eliminating the
reverse hypothesis, since effects do not precede their causes63. However, the strategy
outlined above is one of eliminative induction (Bird, 2011, 242), and therefore, in order to
establish a clear causal relationship, one must also eliminate (C); that is, one must rule out
every possible alternative explanation for the observed correlation between E and O.
Consider the following observational study:
In order to test whether drinking caffeine causes bowel cancer, a case-control study
is conducted in a hospital. Let us suppose the study reveals a higher risk of exposure to large
quantities of caffeine amongst the bowel cancer group, than amongst the control group
(drawn from the general population). Can one infer a causal relationship between drinking
caffeine and bowel cancer? There are good reasons to think not, since many alternative
explanations cannot be ruled out. Those who drink excessive quantities of caffeine might
work longer hours, be more stressed, work in hospitals, smoke more, drink more alcohol,
eat more red meat, and so on – these are all possible explanations for the higher incidence
rate. Furthermore, even if we seek to rule out every possible alternative explanation, how
could we have done this? Perhaps there are common causes we haven’t thought of. The
6363
I am sure some metaphysicians will question this (anyone who believes time travel is possible, for
example), but in the context of medical research, this is a fair assumption!
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
96
application of Hill’s criteria to an observational study, then, cannot eliminate C, since these
criteria ‘fail to exclude the common cause hypotheses in a systematic way’ (Bird, 2001, 244).
Examples like this highlight the fallibility of the epidemiologist’s causal inferences,
particularly in the case of observational studies64. In the next section, however, I investigate
what has become a topical discussion in the philosophy of epidemiology (Broadbent 2015) –
a theoretical means of inferring causal relationships in epidemiology, based on
counterfactuals.
2. The epidemiologist’s potential outcomes approach
In this section I present the epidemiologist’s counterfactual, or potential outcomes
approach (POA) to causation - thismethod of identifying causal relationships shares qualities
with the counterfactual conceptions found in traditional analytic philosophy (Lewis 1973), as
the methodology presupposes that the effect of an exposure is measured relative to some
contrary-to-fact condition65. In 2.1 I first outline the general POAstrategy, and how it relates
to the counterfactual strategies most analytic philosophers are acquainted with; in 2.2 I
argue that although there are similarities, the links between the philosophical and
epidemiological counterfactual approaches to identifying causal relationships are somewhat
tenuous. I then demonstrate in 2.3 that for some common parameters (such as those
presented in Cartwright’s 2007 paper on RCTs), if the POA does succeed as a means of
identifying relevant causal relationships, it does so using a frequentist probability raising
64
Howick (2004) and Bird (2011) argue that experimental studies, like RCTs, are far less vulnerable to inferring
spurious causal relationships.
65
When asking whether A caused B, Lewis asks us to consider a world in which everything is the same, aside
from the fact that A does not occur (contrary-to-fact) – A causes B if, in this world, B does not occur (see
section 2.3 of this chapter).
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
97
account. In 3.1 and 3.2I present Broadbent’s characterisation of Hernan and Taubman’sPOA,
and outline why he believes thePOA involves both circular and false hypotheses – I then
attempt to refute these objections in 3.3.
I conclude that the POA strategy is not designed to identify all ‘in principle’
(Woodward 2004) causes of disease. It is a tool designed to measure the effect of
practicableinterventions, and as such, does not deal with nonmanipulable causes (which are
very often of fundamental importance to both epidemiological studies and clinical
medicine). Given that the POA theoretical framework cannot accommodate nonmanipulable
causes, in section 4 I propose that in some circumstances, Rothman’s sufficient-cause model
is more suitable for the epidemiologist.
2.1 The potential outcomes approach methodology
The POA is concerned with many ‘potential outcomes’ (whereby outcomes can be a number
of variables, including incidence rates, life expectancy, and so on), the values of which are
determined through (a) actual group studies, and (b) estimates of the outcomes of
counterfactual studies; that is, not only whether or not the actual outcome occurs given the
non-occurrence of the actual exposure, but specific data concerning outcomes under a
number of possible contrary-to-fact exposures. The strategy thus runs roughly as follows:
There are a number of possible actions, only one of which is actual. Take the actual
action of an individual or population to be x0, and all alternative exposures to that individual
or population to be uniquely specified: x1, x2, x3, and so on.
(i) Take Oto be a measure of outcome, and O(x0) to be the outcome of the observable
event x0. O(xn) for all values of n except 0 are unobservable – they are the
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
98
counterfactual outcomes (that is, what would have occurred were x1, or x2, or...
to have occurred). All outcomes O(xn) are potential outcomes (including the
actual outcome).
(ii) Compare the actual outcome O(x0) with any one counterfactual outcome O(xc), to
measure the value of x0 versus O(xc)
Outcomes other than O(x0) are contrary-to-fact and thus unobservable, but reasonable
estimation (assuming this is possible) of these values admits of numerous effect-measures
(for example, if one takes x0 to be ‘immunised against yellow fever’, and x1 to be ‘not
immunised against yellow fever’, one can calculate the risk of yellow fever ratio due to nonimmunisation versus immunisation by O(x1)/O(x0)).
The strategy employed implies that it is meaningless to assert that ‘non-immunisation is
a cause simpliciter of yellow-fever’ – one must first determine which intervention/action it is
relative to: one must assert that ‘non-immunisation causes yellow fever versus
immunisation’. That one must always assert what an exposure is relative to, leads to some
surprising consequences. For example, unprotected penile-vaginal sex (where H.I.V. is
prevalent) is a cause of H.I.V. versus abstinence; but it is not a cause of H.I.V. versus
unprotected anal sex (where H.I.V. is prevalent) – indeed, relative to unprotected anal sex,
unprotected penile-vaginal sex reduces risk of H.I.V. Intuitive or not, this strategy is
important.
2.2The potential outcomes approach and the Lewisian conception
Epidemiologists are usually concerned with studies confirming or falsifying general causal
claims such as ‘smoking causes cancer’, or ‘the Plasmodium vivax malaria transmission is
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
99
more resilient to interruption than other forms of malaria’ (Mendiset al, 2009) - not so for
the most influential philosophers espousing counterfactual conceptions of causation; they
are more often interested in well-specified token events (eg. ‘Joe’s contracting malaria’) –
let us call this approach the ‘Lewisian conception66’. Given that the epidemiologist is
concerned with general causation, and the Lewisian with singular causation, the ties
between Hume and Lewis’s counterfactual conditional accounts and the epidemiologist’s
POAare tenuous in many respects. Nevertheless, proponents of the POAadopt a similar
strategy to the Lewisian, in that they judge causal effects based not only on the actual
outcomes of a patient’s actual exposure, but the potential outcomes of alternative,
unrealised exposures on the same patient(s) - assigning the phrase ‘counterfactual’ account
thus seems appropriate to both the Lewisian conception, and the POA.
As we saw in chapter 3, the Lewisian considers causation to be a matter of
counterfactual dependence, whereby counterfactuals are evaluated using his ‘possible
worlds’ ontology. This conception is of little use to the epidemiologist, of course, since
moving from just a single causal inference (of the kind identified by the Lewisian method),
to more general causal claims, is unjustified. Suppose, for example, that unbeknownst to
Bob, Bob had always had a black mamba living under his bed. Bob habitually got out of bed
quietly in the mornings, and consequently never woke the snake. Although it is true that
66
Hume’s work is often seen as the origin of counterfactual conditional accounts of causation – it may seem
surprising, then, that I refer to the view as ‘Lewisian’; I call the philosophical (as opposed to epidemiological)
counterfactual analysis of causation the ‘Lewisian counterfactual conception’ for two reasons: first, because
Lewis’s account is (unsurprisingly, given a couple of centuries’ progress in philosophy) more sophisticated and
less open to counterexamples than Hume’s, but second, and more importantly, because Hume provides a
regularity approach to causation (for which, it is probably fair to say, he is more often associated with by
analytic philosophers) in addition to his counterfactual approach, not at all grounded by counterfactuals (if
anything, counterfactuals are grounded by regularities). More on regularity approaches to causation in later
sections.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
100
getting out of bed quietly was a cause of Bob living a long and happy life, the general causal
claim ‘getting out of bed quietly increases life expectancy’ is clearly false.
2.3Reichenbach and the potential outcomes approach
Identifying causes in the way proposed by the POA(that is, via risk-parameter measurement
(and comparison)), on the other hand,can be a notational variant of the frequentist
probability-raising account of cause, similar to that proposed by Hans Reichenbach67 (1956).
To establish a causal relationship using risk-parameters, where x is a specified
population, one measures whether an effect is more probable given a well-defined
intervention by calculating the outcome O(x0) (say, risk of morbidity), and the
counterfactual outcome of the same parameter O(x1), and comparing the two. Risk is
equivalent to frequentist probability (the number of cases/population at the start of the
study), so if the risk of morbidity given x0 is higher than the risk given x1, that is, P(D|x0) >
P(D|x1)68, according to the Reichenbach probability-raising account of cause, x0 is to be
deemed a cause of diseaseD (given the nature of the POA, one must specify that x0 is a
cause relative to x1, of course). Note that, in the case of observational studies at least, this
simplified model will present ‘spurious correlations’ as causal relationships, as a result of
confounding factors (as highlighted by the ‘alcohol causes lung cancer’ example).
Proponents of the POAtake this confounders problem to be adequately dealt with by
randomisation (and where randomisation is not possible, similar methodological strategies)
– the problem is thus a methodological one, to be dealt with through careful study-design.
67
RCTs certainly employ this kind of approach to causation.
68
To be read: The probability/risk of disease given x0 is greater than the probability/risk of disease given x0
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
101
This short, one sentence response will hardly satisfy most readers, but the confounders
problem is one that plagues epidemiological methods of all kinds.
BothReichenbachand Lewis present their views as accounts of singular causation.
They are clearly distinct analyses of causation, but(at least with respect to risk parameters)
the POA is as similar to the Reichenbach view as it is to Lewis’s account. The similarity to the
Lewisian view rests on the use of counterfactuals, insofar as effect-measurements require
statistics for outcomes under hypothetical scenarios. The similarity to the Reichenbach view
rests on effect-measurements being made through the epidemiological equivalent to
relative-probabilities; that is, by comparing the value of some parameter that plays the
‘probability of the effect given the (proposed) cause’-role, and the value of the same
parameter equivalent to the ‘probability of the effect without the (proposed) cause’. Note,
however, that not all parameters used by epidemiologists are equivalent to frequentist
probability, so not all effect-measures made possible by the POA will be notational variants
of the probability-raising account. One can measure the effects of smoking in terms of life
expectancy given smoking versus life expectancy given non-smoking, for example, but life
expectancy is not a risk-measurement expressed in terms of frequentist probabilities.
I do not invite the reader to identify the numerous counterexamples to my simplistic
exposition of Reichenbach’s probability-raising conception, or that of Lewis’s counterfactual
theory (largely as more comprehensive expositions of both theses are more defensible)69.
The expositions I have provided of both views, however, are sufficiently close to the more
refined versions to establish the following claims: (i) that the Lewisian conception permits
69
If the reader is interested in these philosophical issues she should consultReichenbach’s (1956), and Lewis’s
(1973 and 2000) work directly. More detailed expositions of the POAcan be found in Rothman, Greenland and
Lash (2008, chapter 4), and Greenland (2005)
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
102
identifying cases of singular causation, whereas the epidemiologistmust, just for pragmatic
purposes, identify cases of general causation; (ii) the Lewisian conception and the POA are
alike insofar as both involve consideration of contrary-to-fact suppositions, as well as one
observable outcome; (iii) the POAonly provides effect-measures due to the exposure versus
an alternative specified (counterfactual) condition; and (iv) that in essence, when using riskmeasurements, a POAanalysis of cause can be viewed as a probability-raising account
similar to the Reichenbach’s, since risk in epidemiology is a notational variant of
Reichenbach’s frequentist probability; but again, the POAis differs insofar as itis concerned
with general causal statements, and there are alternative parameters epidemiologists use
(such as life expectancy) which do not fit well with a probability-raising conception.
3.1Hernan and Taubman’s potential outcomes approach
Epidemiology is a prescriptive discipline,directed at improving public health through group
studies. Given its prescriptive nature, prima facie epidemiologists need only be concerned
with manipulable conditions – if one cannot prevent a condition being satisfied, and one
cannot interfere with a condition as a form of treatment, one may as well consider it not a
cause. To illustrate: it would be odd to consider ‘being male’ a cause of testicular cancer,
even though only males contract the disease (in Mackie’s eyes, being male would simply be
a part of the ‘causal field’70). It might be sensible, however, to study the relationship
between testosterone levels and testicular cancer, as testosterone levels can be interfered
with). On the face of it, then, information concerning causal relationships between
manipulable conditions (like testosterone levels) and diseases is, unlike information about
nonmanipulable conditions like ‘being male’, useful when deciding public health policy. It is
70
See chapter 3
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
103
this thought that drives Hernan and Taubman (2008) to assert that one cannot make claims
about causal effects (within the epidemiological context) without specifying at least one
well-defined intervention.
Here I criticise what I deem to be a naïve interpretation of the POA,by highlighting
three objections raised in Alex Broadbent’s (2015) paper on causation and prediction in
epidemiology. Broadbent considers Hernan and Taubman’s (two POAtheorists) 2008 paper
on obesity, in which they claim that no causal questions about the effects of obesity are
meaningful. Broadbent takes this paper (along with significant additional evidence) to
capture the essence of thePOA, identifying four theses he deems advocates of thePOAto be
committed to: one semantic, one metaphysical, one pragmatic, and one epistemic.
3.2 Broadbent’s characterisation of Hernan and Taubmen’s potential outcomes approach
The semantic thesis states that our pretheoretic intuitions about causation should be
ignored when developing an account of causation within epidemiology, acknowledging that
it is the pragmatic consequences of the causal-conception that marks its quality. The
semantic thesis thus captures the ‘improving public health’ prescriptive nature of thePOA
(so there is nothing obviously incoherent about it).
The metaphysical thesis states that ‘…at least sometimes, causes are differencemakers. X is a difference-maker for Y if and only if, had X had been absent or different, then
Y would have been absent or different’ (p. 5). This is a somewhat weaker version of
Hume’sconditional definition of cause, in which he state’s ‘if the first object had not been,
the second never existed’ (Hume, Enquiry, Sect. VII, Pt. II71), since causes do not (necessarily)
71
Oxford University Press (1999) p. 146
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
104
always make a difference under the POAcriterion - although smoking causes cancer, it might
be the case that an individual would have developed cancer regardless of whether or not
she smoked (perhaps due to some genetic predisposition). The weakening of the
counterfactual thesis maintains the POA’scommitments, without falling foul to obvious
objectionsagainst anything stronger. This weakening is necessary due to the move from
singular to general causation (but there is no need to explore this any further).
The pragmatic thesis states that ‘…the only difference-makers that epidemiology
needs to care about are those that are humanly manipulable.’ (p. 5) This aspect is of course
closely linked with the semantic thesis, as prima facie for epidemiologists, usefulness is
largely determined by what can humanly be changed. Unlike the semantic and metaphysical
theses, Broadbent challenges the pragmatic thesis:
According to the pragmatic thesis,in order to know whether some condition is causal
one must first know whether or not that condition can be (humanly) manipulated. This
knowledge, says Broadbent, can only come from one of two sources. The first is background
scientific knowledge, but this always involves causes without humanly manipulable
interventions. The second is via direct empirical evidence: one attempts to manipulate some
condition under investigation, and where one discovers that it does not correspond to a
well-defined intervention, one declares it non-causal. Broadbent argues that in order to
conduct such an experiment, one needs to have already identified a well-defined
intervention, and since one does not know in advance whether an intervention is welldefined, the pragmatic thesis is flagrantly circular.
Finally, the epistemic thesis states that ‘…causal knowledge yields predictive
knowledge…. Enabling prediction under hypothetical scenarios is the mark of causal
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
105
knowledge’ (pp. 5-6). According to Broadbent this epistemic thesis can also be refuted, this
time simply on the grounds that it is demonstrably false. It is undeniable that ‘smoking
causes cancer’ is a useful causal claim, yet this knowledge has yielded many false predictions
– e.g. smoking low-tar cigarettes, or taking shallower puffs, is better for smokers than
smoking standard cigarettes in the normal way.
In the following section I argue that although Broadbent’s criticism of the Hernan and
Taubman (and in particular the pragmatic thesis) is understandable (for reasons I will
demonstrate later in this chapter), the thesis epidemiologists are predominantly interested
in, the semantic thesis, warrants rethinking the pragmatic and epistemic theses such that
they avoid the circularity and falsity objections. In the following sections I suggest that
applying the Popperian scientific method resolves Broadbent’s worries; furthermore, I show
that causal knowledge obtained via the POA does in fact yield at least some predictive
knowledge.
3.3 Diffusing Broadbent - A Popperian Take on the Potential Outcomes Approach
Broadbent’s objection to the pragmatic thesis is predicated on Hernan’s inability to answer
‘how can the epidemiologist know which conditions are humanly manipulable?’ without
succumbing to a vicious circle. As we saw, the two possible approaches to answering the
question (as identified by Broadbent) are (i) existing scientific background knowledge – but
this fails because it involves ‘a large quantity of causes that do not correspond to humanly
manipulable interventions’ (p. 6), and (ii) empirical investigation – but this fails as ‘for good
empirical enquiry, you need already to have identified a well-defined intervention’ (p. 6).
However, although Broadbent is right to claim only existing scientific knowledge and
empirical enquiry can provide the information, this is not problematic.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
106
The paper Broadbent focuses on questions whether obesity shortens life (Hernan
and Taubman, 2008). Hernan and Taubman conclude that obesity cannot be considered a
cause because it does not correspond to a well-defined intervention – demonstrable, in this
case, because there are many ways to lose weight72, each of which has a distinct effect on
mortality; that is, empirical studies designed to measure death attributable to obesity give
different values depending on the mode of intervention. Given that there is no well-defined
intervention, obesity is not a condition suitable for causal effect measurement.
Although the Hernan/Taubman paper is prima facie convincing regarding the noncausal status of obesity, Broadbent points out that ‘…if we [discover whether something
corresponds to a well-specified73 intervention] by empirical investigation, then we cannot
have decided in advance74 whether an intervention is well-specified… Thus we cannot make
the question of whether our empirical investigation is a good one depend on advance
knowledge of whether the putative causes we are contemplating correspond to wellspecified interventions’ (p. 7).
Although Broadbent’s argument is sound, such a strategy is only problematic under
certain assumptions - the key question being: ‘to what extent must we know that an
intervention is well-defined prior to investigation?’ Suppose that instead of reading the POA
to assume causes require ‘confirmed’ well-defined interventions, one is more charitable,
and takes Hernan to assume epidemiologists adopt a broadly Popperian strategy; that is,
rather than taking an hypothesis to be unequivocally proven, they use our best scientific
72
Different diets, different exercise routines, smoking, and so on.
73
‘ Well-specified’ and ‘well-defined’ should be taken as synonyms.
74
My emphasis.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
107
theories, and analogies with similar cases of better-corroborated well-defined interventions,
to determine (with a well-educated guess) which potential causes have well-defined
interventions. Of course, Broadbent is right in his assertion that observation must come
subsequent to theory, but this passage from Popper’s 1972 suggests this fact might not be
problematic, but quite usual in the sciences.
‘The belief that science proceeds from observation to theory is still so widely and so
firmly held that my denial of it is often met with incredulity… But in fact the belief
that we can start with pure observation alone, without anything in the nature of
theory, is absurd… Observation is always selective. It needs a chosen object, a
definite task, an interest, a point of view, a problem. And its description presupposes
a descriptive language, with property words; it presupposes similarity and
classification, which in turn presupposes interests, points of view, and problems.’
(Popper, 1972, 46)
On a Popperian, falsificationist view, the epidemiologist can assert that an empirical
investigation is a good one if our best scientific theories identify the relevant intervention as
‘well-defined’ in advance of the empirical investigation, whilst maintaining an appropriate
level of scepticism (and attempting to falsify the hypothesis through empirical
investigation).
Suppose, for example, we did not know that smoking could lead to weight loss, and
our best guess is that the only way of losing weight is to eat healthily. Although we could not
know for sure that obesity (the proposed cause of, say, heart disease) corresponds to a welldefined intervention (which, it later turns out, it doesn’t), that is not to say we were not
justified in treating it as if it does, given our ‘knowledge’ at the time (later, this hypothesis
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
108
would be falsified, but that’s OK). One is not asserting that obesity definitely corresponds to
a well-defined intervention, only that, given our current understanding of similar cases, it is
likely that it does – there is no circularity here. Of course, once science reveals that obesity
does not have a well-defined intervention, it is no longer a candidate for a cause of disease,
but that does not mean the initial investigation was a bad one. A broadly Popperian attitude
to scientific investigation, then, diffuses Broadbent’s circularity objection to the POA.
What of the falsity objection to the epistemic thesis: that ‘Enabling prediction under
hypothetical scenarios is the mark of causal knowledge’ (p. 6), yet the POAconception of
causation regularly yields false predictions? I do not think one should be concerned by this
objection, either. On the face of it, the objection is only valid if ‘enabling prediction under
hypothetical scenarios’ requires the predictions to come true – of course, perfectly justified
predictions can (and often do) turn out false, so this cannot be what Broadbent has in mind.
Indeed, the objection is clarified in a footnote. Referring to the false predictions made about
what happens when tar in cigarettes is lowered, he states:
‘Strictly speaking… I should say that these predictions were not only false but also
bad in some more substantive way. In particular… the implementation of causal
knowledge in making those predictions did not constitute checking how one might
be wrong, and thus did not by itself amount to a good prediction activity’ (2015, 7)
Although Broadbent is right to highlight the importance of the epistemic thesis in the POA,
his refutation relies on his own conditions for what it is to be a good prediction activity.
Clearly, true predictions can fall out of a poor prediction activities (wild guesses can turn out
to be true), and false predictions can fall out of good prediction activities. However, if
Broadbent’s objection is to hold any weight, he would have to deny the following
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
109
statement: ‘If, following a population of fast food eaters, one knows via the POAthat the
counterfactual outcome O(x healthy diet) is a high life expectancy versus actual outcome
O(x fast food diet), then one can predict (well) that if a population changes their eating
habits from a fast food diet to eating healthily, their life expectancy increases.’ - the causal
statement alone does not consider alternative outcomes to the counterfactual scenario, and
thus according to Broadbent’s criteria, the prediction activity is a bad one – a
counterintuitive conclusion indeed (in fact, the relation between the causal and predictive
claims looks to be a matter of logical entailment). Let us consider Broadbent’s example to
see why this comes about:
‘As a descriptive claim, the epistemic thesis is clearly false, as there are plenty of
cases where we have what we might call ‘causal knowledge’ but make bad
predictions about what will happen under hypothetical suppositions. Smoking causes
lung cancer, yet predictions based on the knowledge about what would happen if
the amount of tar in cigarettes were lowered, or if smokers took shallower puffs
were mistaken’ (p. 7).
Take A to be ‘smoking with normal puffs’; B to be ‘lung cancer’; and C to be ‘smoking with
shallow puffs’. The structure of the inference above thus runs as follows: we know that ‘A
raises the probability of B’ (A causes lung cancer), therefore we can justifiably infer ‘C raises
the probability of not-B’ (if I take shallower puffs then I am less likely to get lung cancer).
This is quite clearly a fallacious move, since A and C are distinct activities (despite both
involving smoking, C is not the negation of A). No sensible advocate of ‘causal knowledge
yields predictive knowledge’ would think this inference infallible. Good prediction activity,
when seen from a causal perspective, requires the right causal knowledge. We know that
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
110
smoking causes cancer versus non-smoking, and, as required by the epistemic thesis, this
yields predictive knowledge. Specifically, it enables predictions like ‘if this population stops
smoking, they are less likely to get lung cancer75,’ but there is a limited amount of predictive
knowledge that any one causal claim can yield. The predictive knowledge is, in fact,
sometimes determinately specified by the counterfactual scenarios considered in the
POAstudy: If smoking causes cancer versus not smoking, then one can predict stopping
smoking will reduce risk of lung cancer. If taking ibuprofen reduces headache versus taking
placebos, then one can predict taking the ibuprofen over the placebo will reduce the
headache. The move from ‘ibuprofen reduces headache versus placebo’ to ‘my headache
will go away faster if I take ibuprofen, than if I take paracetamol’, however, is a bad
prediction activity, as the POAcausal claim (remember that POAcausal claims are
meaningless without specifying a contrast) makes no reference to any counterfactual
scenario involving paracetamol. The POAdoes enable prediction (albeit without
guaranteeing the prediction will be true), and given the very nature of POAcausal claims,
POA causal knowledge alwaysyields some predictive knowledge. Of course, the strategy the
POAemploysmakes it possible for one to be incorrect about effect-measures, as estimations
of outcomes under counterfactual scenarios can be very wrong (one might, for example, be
completely unaware of some condition that would affect the outcome of the counterfactual
scenario), but this is a problem with the POA’s overall strategy, rather than the epistemic
thesis. Hernan and Taubman should not, then, be overly worried by Broadbent’s charge.
4.1 The Importance of Nonmanipulable Causes
75
This requires the accuracy of our counterfactual suppositions in the POAprocess of identifying causes, of
course, but one must assume this.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
111
Nancy Cartwright, in her 2010 PSA presidential address criticising RCTs,demonstrated that in
practice, unjustified extrapolations from ‘intervention X works somewhere’ to the claim that
the ‘intervention X works in general’ are regularly made. She highlights two instances of
nutritional intervention funded by The World Bank. The first was conducted in Tamil Nadu
state, India, and comprised providing food supplements, health measures and educating
pregnant mothers about how to better nourish their children. The project was very
successful, with malnutrition dropping significantly in the region. The same programme was
then implemented in Bangladesh - an area with similar problems. Given the success in Tamil
Nadu, a POAstudy would suggest a similar drop in malnourished children elsewhere, but in
Bangladesh there was no such drop. So what explains the failure of the inferred general
causal claim? Two reasons were proposed: first, ‘food supplied by the project was used not
as a supplement but as a substitute, with the usual food allocation for that child passing to
another member of the family’ (Cartwright 2010, 13), and second, that ‘the program
targeted the mothers of young children. But [in Bangladesh] mothers are frequently not the
decision makers… with respect to the health and nutrition of their children’ (pg 13).
In Bangladesh, then, the background conditions were such that the intervention
funded by The World Bank was insufficient to reduce malnutrition, since even with the extra
food and education, conditions together sufficient for malnourishment were in place. On
the other hand, in Tamil Nadu educating young mothers removed a non-redundant part of
the conditions sufficient for their children’s malnourishment (and in many cases, there were
no other jointly sufficient causes), so the program was successful. This study implies that in
certain circumstances, an alternative to the POA causal model, like the conception provided
by Rothman below, must be applied.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
112
As one can see through the example above, an effect is often not determined by a
single causal factor – indeed, usually there are several states of affairs that are only together
sufficient for the effect. This is captured by the sufficient-cause model (Rothman 1976)
outlined in the previous chapter, according to which a sufficient cause is represented by a
pie-chart76.
Figure 8 below represents the pie chart for child-nutrition – it shows that the
sufficient cause for a well-nourished child includes sufficient food, a well-educated nutrition
decision-maker, the correct distribution of the food, and ‘other’ unspecified non-redundant
conditions. Each segment of the chart is a necessary condition for good child-nutrition. Prior
to the intervention, in both Tamil Nadu and Bangladesh segments of the pie were missing.
In Tamil Nadu there was insufficient food and the decision maker (the mother) was
insufficiently educated in child-nutrition. The project funded by The World Bank
implemented interventions such that the sufficient cause was completed and the childmalnutrition levels dropped. In Bangladesh, however, educating the mothers in childnutrition and providing food supplements did not have the desired effect, as only one of the
absent conditions in the sufficient cause was satisfied. First, because educating the mother
did not help with the ‘educated nutritional decision-maker’ condition, and second because
although there the project provided sufficient food, the household food was not being
distributed in the manner intended.
76
See chapter 3, section 3
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
113
Figure 8: Malnutrition’s ‘sufficient-cause’
Perhaps the most important lesson here is that a sufficient-cause is unlikely to
comprise only humanly manipulable states of affairs - it is highly unlikely, for example that
an epidemiologist can, in practice, change the decision-making processes of an entire
culture. Even if this were possible, amongst the ‘other non-redundant conditions’ will be
factors that are nonmanipulable: the price of healthy food, whether or not there were
floods or droughts, and so on. The epidemiologist must, when making decisions about
studies to conduct (and what public health policies to suggest), take at least some
nonmanipulable causes into consideration.
4.2 Nonmanipulable Causes, epidemiology, and clinical decision making
Cartwright’s example highlights the fact that outcomes and interventions are not the only
important factors in assessing the kind of counterfactual epidemiologists are interested in,
as additional conditions shape the results of studies. In the obesity example, Hernan and
Taubman argue that one should not consider obesity a cause of increased morbidity since
obesity does not have a well-defined intervention – ultimately, this is because ill-defined
interventions make assessing the counterfactual ‘if she were not obese, she would have
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
114
lived longer’ impossible, since different interventions produce different truth-values. But the
same is true of any intervention in which nonmanipulable factors affect the truth-value of
counterfactuals, when these are not well-specified. For example, two people may have the
same deadly bacterial infection, but one cannot necessarily state that ‘if the patient ingests
penicillin her risk of death is lower’ when important, nonmanipulable conditions like
penicillin allergies have been ignored. Not only do the interventions need to be wellspecified, but nonmanipulable background conditions, too - they ‘make a difference’. The
POAmay be a thesis about interventions; a thesis driven by what epidemiologists and
medical practitioners can actually do to prevent or treat diseases, but there is no
questioning the causal relevance of nonmanipulable factors in both epidemiology and
clinical medicine.
The POA is not a suitable candidate for identifying all causes of disease, even within
an epidemiological framework. The structure of the approach provides effect-measures of
manipulable conditions, and of course the effects being measured have causes – namely,
the well-defined interventions 0,1,2…n, but manipulable conditions are not the only
relevant causes in epidemiology. Of course, that the POA provides a means of measuring the
effect of one intervention versus another, and does not accommodate nonmanipulable
states of affairs, does not rule nonmanipulable states of affairs non-causal. All one can
conclude from the method outlined in this chapter, is that the POA cannot measure the
effect of nonmanipulable causes.
This conclusion does not suggest that allnonmanipulable but nonredundant states of
affairs shouldbe considered causes by those involved in medical research. Epidemiology is a
prescriptive discipline, and thus only those events/states of affairs that it is useful to call
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
115
causes should be considered causes. Nonetheless, there are many cases in which it is
necessary to take nonmanipulable causes seriously, such as the case of malnutrition in Tamil
Nadu and Bangladesh. There I used Rothman’s sufficient-cause model to map the relevant
causal factors – a suitable candidate, I think. But this would not be a suitable causal-model
for measuring the causal effect of smoking on life expectancy. How one can best model
causation in epidemiology, then, looks to change depending on context.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
116
Conclusion:
Context Dependency is Rife
In this book I have attempted to provide analyses of the concepts and methods used by the
medical profession. Medicine is ultimately a goal-directed discipline: to improve the health
of the general public. This is true whether one is an epidemiologist discovering causes and
effective interventions, a pathologist examining tissues for diagnostic purposes, or a
clinician attempting to improve the wellbeing of her patients.
There are many different concepts of disease, and many different concepts of
causation. Some of these are entirely unsuccessful, but some are successful in some
contexts but not in others. A value-free concept of disease is useless for deciding whetheror
not an individual should be legally responsible for her actions on health grounds, since this
requires an assessment of her state of wellbeing – an inherently value-laden notion. The
pathologist does not make value judgements, however. The pathologist examines organs,
tissues and cells to determine whether or not they are performing their natural function to
the required standard, and if not, why not – she does not need to consider whether the
patient deems whatever condition he has to be a bad thing. In medicine, the ‘right’ concept
of disease is that which is most useful at the time – it is entirely context dependent.
The same is true of concepts of causation. Notions of causation are of course
fundamental to epidemiology. Some kinds of trial presuppose a very particular concept of
causation. Cartwright states that RCTs are seen as the gold standard because they imply
their causal conclusions – but this can only be asserted if one presupposes a probabilityraising account of causation. The POA is great for measuring the effects of
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
117
manipulableinterventions, but the epidemiologist cannot forget that manipulable
interventions are not the only type of cause. She must employ a completely different
concept of causation, such as the sufficient-cause model, if she is to accommodate
situations in which nonmanipulable factors play a role – that model, however, is no good for
measuring causal effects.
Thus, the concept of disease that is useful to the clinician is different to that
pertaining to pathology, and both are different again from that which is appropriate to the
epidemiologist.
Bibliography
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
118
Ananth, M. 2008. In Defense of an Evolutionary Concept of Health: Natures, Norms and Human
Biology. Aldershot: Ashgate Publishing Limited
Bird, A. 2007. Nature’s Metaphysics: Laws and Properties. Oxford: Clarendon Press.
2011. The epistemological function of Hill's criteria. Preventive Medicine, 53(4-5): 242-5.
Broadbent, A. 2008. The Difference between Cause and ConditionProceedings of the Aristotelian
Society, 108: 355-64
2009. Causation and Models of Disease in Epidemiology. Studies in History and Philosophy of
Biological and Biomedical Sciences, 40(4): 302-311.
2012. Causes of Causes. Philosophical Studies, 158(3): 457-476.
2013. The Philosophy of Epidemiology. Palgrave Macmillan.
2015. Causation and Prediction in Epidemiology: A Guide to the Methodological Revolution,
Studies in History and Philosophy Biological and Biomedical Sciences, S1369-8486(15):
00084-9.
Boorse, C. 1975. On the Distinction between Disease and Illness. Philosophy and Public Affairs, 5(1):
49-68.
1976. What a Theory of Mental Health Should be. Journal for the Theory of Social Behavior,
6(1): 61-84.
1977. Health as a Theoretical Concept. Philosophy of Science, 44(4): 542-573.
1997. A Rebuttal on Health. In J, M. Humber and R, F. Almeder (eds). What is Disease?
Totowa: Humana Press.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
119
2002. A Rebuttal on Functions. In Functions, eds. A. Ariew, R. Cummins, and M. Perlman, 63112. New York: Oxford
2014. A Second Rebuttal on Health. Journal of Medicine and Philosophy, 39(6): 683-724.
Cartwright, N. 2007. Are RCTs the Gold Standard? BioSocieties, 2(01): 11-20.
2012. Will this Policy work for You? Predicting Effectiveness Better: How Philosophy Helps.
Philosophy of Science, 79(5): 973-989.
Collins, J., Hall, N., & Paul, L.A. 2004. Causation and Counterfactuals. Cambridge: The MIT Press.
Cooper, R. 2002. Disease.Studies in History and Philosophy of Science, 33(2): 263-282.
Dawkins, R. 1976. The Selfish Gene. Oxford: Oxford University Press.
DeVito, S. 2000. On the value-neutrality of the concepts of health and disease: unto the breach
again. Journal of Medicine and Philosophy, 25(5): 539-67.
Eke, P.I., Dye, B.A., Wei, L., Thornton-Evans, G.O. &Genco, R.J. 2012. Prevalence of Periodontitis in
Adults in the United of States: 2009 and 2010. J Dent Res, 91(10): 914-20.
Ereshefsky, M. 2009. Defining ‘Heath’ and ‘Disease’. Studies in History and Philosophy of Biological
and Biomedical Sciences, 40(3): 221-7.
Ellis, B. 2001. Scientific Essentialism. Cambridge: Cambridge University Press.
2009. The Metaphysics of Scientific Realism. Durham: Acumen Publishing.
Ellis, B. &Lierse, C. 1994. Dispositional Essentialism. Australasian Journal of Philosophy, 71(1): 2745.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
120
Engelhardt, H.T. Jr. 1976. Ideology and Etiology. Journal of Medicine and Philosophy, 1(3): 256-68.
Eriksen, T et al. 2013. At the borders of Medical Reasoning: Aetiological challenges of medically
unexplained symptoms. Philosophy Ethics and Humanities in Medicine, 8: 11.
Fisher, R.A. 1925. Statistical methods for research workers. Edinburgh: Oliver and Boyd.
Flew, A. 1973. Crime or Disease? London: The Macmillan Press Ltd.
Geach, P. 1972. Logic Matters. Berkely: University of California Press.
Giroux, E. 2015. Epidemiology and the bio-statistical theory of disease: a challenging perspective.
Theoretical Medicine and Bioethics, 36(3): 175-95.
Guerrero, D.J. 2010. On a Naturalist Theory of Health: A Critique. Studies in History of Philosophy of
Biological and Biomedical Sciences, 41(3): 272-8.
Hall, N. 2004. Two Concepts of Causation. In Collins, Hall, & Paul, Causation and Counterfactuals:
225-276. Cambridge: The MIT Press.
2000. Causation and the Price of Transitivity. The Journal of Philosophy, 97(4): 198-222.
Hausman, D.M. 1996. Causation and Counterfactual Dependence Reconsidered. Nous, 30(1): 55-74.
2012. Health, Wellbeing, and Measuring the Burden of Disease. Population Health Metrics.
10(13): 1-7
Heil, J. 2012. The Universe as We Find It. Oxford: Oxford University Press.
Hek, K. et al. 2013. A Genome-wide Association Study of Depressive Symptoms. Biological
Psychiatry,73(7): 667-78.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
121
Hernan, M.A. 2005. Invited Commentary: Hypothetical Interventions to Define Causal Effects Afterthought of Prerequisite? American Journal of Epidemiology, 162 (7): 618-620.
Hernan, M. A, and Taubman, S. L. 2008. Does Obesity Shorten Life? The Importance of Well-Defined
Interventions to Answer Causal Questions. International Journal of Obesity, 32. S8-S14
Hesslow, G. 1993. Do we need a Concept of Disease? Theoretical Medicine, 14(1): 1-14.
Hill, A.B. 1965. The Environment and Disease: Association or Causation? Proceedings of the Royal
Society of Medicine, 58: 295-300.
Hume, D. 1999. An Enquiry Concerning Human Understanding. Oxford: Oxford University Press.
Johnson, M. 1992. How to Speak of the Colors. Philosophical Studies, 68(3): 221-263.
Kerry, R. et al. 2012. Causation and Evidence-based Practice: An Ontological Review. Journal of
Evaluation in Clinical Practice, 18(5):1006-12.
Kingma, E. 2007. What is it to be Healthy? Analysis, 67(294): 128-33.
2010. Paracetamol, Poison and Polio: Why Boorse’s Account of Function Fails to Distinguish
Health and Disease. British Journal for the Philosophy of Science, 61(2): 241-64.
2012. Health and disease: social constructivism as a combination of naturalism and
normativism. In Carel, H., & Cooper, R.V. 2013. Health, illness and disease: philosophical
essays: 37-56. Durham: Acumen Publishers.
2014. Naturalism about Health and Disease: Adding Nuance for Progress. The Journal of
Medicine and Philosophy, 39(6): 590-608.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
122
Kirk, R. E. 1996. Practical significance: A concept whose time has come. Educational and
Psychological Measurement, 56: 746-759
Kuo, H.K. &Fujise, K. 2011. Human Papillomavirus and Cardiovascular Disease among U.S. Women
in the National Health and Nutrition Examination Survey, 2003 to 2006. Journal of the
American College of Cardiology, 58(19): 2001-6.
Lewis, D. 1973a. Counterfactuals. Oxford: Blackwell Publishers.
1973b. Causation. Journal of Philosophy, 70(17): 556-67.
1979. Counterfactual Dependence and Time’s Arrow. Nous, 13 (4):455-476.
2000. Causation and Influence. Journal of Philosophy, 97(4): 182-197.
Mackie, J.L. 1974. The Cement of the Universe. Oxford: Oxford University Press.
Margolis, J. 1976. The Concept of Disease. The Journal of Medicine and Philosophy, 1 (3): 238-255.
Maslen, C. 2004. Causes, Contrasts and the Nontransitivity of Causation. In Causation and
Counterfactuals (eds.). Collins, N. Hall, N & Paul, L.A: 341-357. Cambridge, MA: MIT Press.
Matsushita, M., Uchida, K., & Okazaki, K. 2007. Role of the appendix in the pathogenesis of
ulcerative colitis. Imflammopharmacology, 15(15): 154-157.
Mill, J. S. 1941. A System of Logic. 8thedn. (reprinted), London.
Millikan, R. G. 1984. Language, Thought, and Other Biological Categories: New Foundations for
Realism. Cambridge, MA: MIT Press.
1989. In Defence of Proper Functions. Philosophy of Science, 56(2): 288-303.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
123
Mumford, S. &Anjum, R. 2011. Getting Causes from Powers. New York: Oxford University Press.
Neander, K. 1991. The teleological notion of ‘function’. Australasian Journal of Philosophy, 69(4):
454-468.
Putnam, H. 1975. Mind, Language and Reality: Philosophical Papers Vol. 2. Cambridge: Cambridge
University Press.
Popper, K. 1972. Conjectures and Refutations: The Growth of Scientific Knowledge 4thedn. London:
Routledge and Kegan Paul.
Rothman, K. J. 1976. Causes Am J Epidemiol. 104: 587-592
Rothman, K. J; Greenland, S; Lash, T. L. 2008. Modern Epidemiology (3rd edition) Lippincott Williams
& Wilkins
Ruse, M. 1987. Biological Species: Natural Kinds, Individuals, or What? The British Journal for the
Philosophy of Science, 38(2): 225-242.
Saracci, R. 2010. Epidemiology. A Very Short Introduction. Oxford: Oxford University Press.
Schaffer, J. 2012. Causal Contextualism in Contrastivism in Philosophy: New Persectivesed. Blaauw
35–63: Routledge
Scheuermann, R. H. Ceusters, W. Smith, B. 2009. Toward an Ontological Treatment of Disease and
Diagnosis. Summit on TranslatBioforma. pp.116-120
Schwartz, P. H. 2007. Natural Selection, Design, and Drawing a Line. Philosophy of Science, 74(3):
364-385
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
124
Smart, B. 2014. On the Classification of Diseases. Theoretical Medicine and Bioethics, 35 (4):251269.
Smith, J.M. 1998. Evolutionary Genetics (2nd edition). Oxford: Oxford University Press.
Wakefield, J. 1992. The Concept of Mental Disorder: On the Boundary between Biological Facts and
Social Values. American Psychologist, 47: 373-388
Whitbeck, C. 1977. Causation in Medicine: The Disease Entity Model. Philosophy of Science, 44(4):
619–637.
Woodward, J. 2004. Counterfactuals and Causal Explanation. International Studies in Philosophy of
Science. 18(1): 41-72
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
125
Index
Ananth
Antidotes
Biological gradient
Biostatistical theory
-value-laden objection
- reference–class objection
-Cambridge change objection
Boorse, C.
Broadbent, A.
Cambridge change
Cartwright, N.
Causes
-
Manipulable
Nonmanipulable
Causation (theories of)
-
Counterfactual
Dispositional
Full-causes
Sufficient-cause model
The vector model
Probability raising
Causal classification of disease
Causal overdetermination
Causal fields
Causal selection
Clinical medicine
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
126
Common diseases
Constructivism
Contrastive studies
Cooper, R.
Diseases (types of)
-
Alzheimer’s disease
Chronic diseases
Dementia
Down syndrome
Parkinson’s disease
Malaria
Mental disorders
Obesity
Sterility
Unwanted pregnancy
Diseases (theories of)
-
Biostatistical
Frequencies and negative consequences
Harmful dysfunction
Harmful function
Pure statistical
Tripartite (Cooper, R.)
Epidemiology
Etiological theory of pathological condition
Evolution
Exposures and outcomes
Falsificationism
Function (natural)
-etiological account
- functional account
- dysfunction
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
127
Guerrero, D. J.
Hill’s criteria
Homosexuality
Interventions
Inusconditions
Kerry
Line-drawing problem
Margolis, J.
Mumford and Anjum
Naturalism
Natural goals
Neander, K.
Normativism
Null hypothesis significance test
Nutrition
Pathology
Possible worlds
Potential outcomes approach
-
Potential outcomes and counterfactuals
Potential outcomes and Reichenbach
Randomised controlled trials
Reference classes
Risk
Rothman, K.
Schwartz, P. H.
Studies
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)
128
-
Experimental
Observational
Contrastive
Vestigial organs
Wakefield, J.
© Benjamin Smart – NOT TO BE CITED WITHOUT PERMISSION (final version forthcoming in Palgrave
Macmillan Pivot Series)